C. A.

HOOKER
SYSTEMATI C REALI SM
1. I NTRODUCTI ON" THREE REVOLUTI ONS I N THE
PHI LOSOPHY OF SCI ENCE
To be a Scientific Realist is not easy - it requires that science be taken
realistically, hence seriously.
Formally, Scientific Realism is a semantical thesis, it is the view that
the intended and proper sense of the theories of science is as literal de-
scriptions of the physical world, as saying what there is and how it behaves.
It is the view that if a scientific theory is in fact true then there is in the
world exactly those entities which the theory says there is, having exactly
those characteristics which the terms of the theory describe them as
having. 1
By Systematic Realism I mean Realism embedded in the repertoire of
supporting philosophical doctrines which together provide a systematic
view of man' s activities, in particular of his scientific activities. For Sci-
entitle Realism, like any other significant philosophical doctrine, does not
stand alone but is linked in an intimate fashion into a variety of other
philosophical and scientific doctrines which are mutually compatible and
mutually supportive (or which, if wrongly chosen, would prove mutually
incompatible and disruptive).
The point of defending Scientific Realism is not simply to insist on a
'right' conception of the semantics of theory. The point of defending it
is beeanse it is part of a coherent account of the entire human animal
and the world in which he lives. According to our best contemporary
theories, humankind evolved slowly from more 'primitive' origins. This
evolution was, for us, most importantly a conceptual evolution in the
broadest sense - man emerged into consciousness of his world. But the
evolution is fundamentally a neurophysiologieal one, an evolution radiat-
ing from the primitive perceptual centre - and its physiological origins
lie even deeper in the skill of survival. Philosophers and scientists aim
at giving a reasonable account of this evolution, which must therefore in-
Synthese 26 (1974) 409-497. All rights Reserved
Copyright 0 1974 by D. Reidel Publishing Company, Dordrecht-Holland
410 ¢. A. HOOKER
dude the physical evolution of the living world, in particular humankind,
the evolution of language and thought, of consciousness, of science, of
values and the general culture. Furthermore it must be a reflective ac-
count, it must explain why the account is reasonable. But knowledge and
reason have themselves evolved, bot h in dynamic interaction with science
and the general culture. Because of the evolution of the former our under-
standing of what reasonableness consists in has itself evolved. So Scientific
Realism then is to play an integral part in the construction of a natural-
istic, evolutionary epistemology, evolving in dynamic interaction with
science itself, and in the construction of a naturalistic, evolutionary ac-
count of mind, of the use of concepts, of language and thought, bot h
linking smoothly into the scientific picture to form the unbroken web of
our world-view. Of course this goal is utopian, our ignorance is t oo great
and epistemological evolution t oo rapid to hope t o complete the task,
but nothing less than this goal provides the proper perspective in which
to view the defence of Scientific Realism. Its defence as an isolated thesis
can only be hopelessly truncated because the stature of the issues is so
reduced - in this form it belongs to boring academia rather than to the
exciting evolution of the planet.
Let us be quite clear about the implications of what I have j ust said for
the exposition of philosophies of science. According to this view, alter-
natives to Realism, e.g. Empiricisms, Conventionalisms, themselves con-
stitute world views, in the sense of containing (albeit sometimes only
implicitly) all of the components mentioned above. And when the Real-
ism/opponents issue is argued back and forth it is the entire world views
of the opposed alternatives that are brought to the attack. Indeed, i f we
agree (counterfactually) to construe Realism strictly as just the semantical
thesis outlined in the second paragraph and to restrict argument to strictly
the semantical level then I do not know of a single argument against (or
distinctively f or) Realism that has ever been advanced (or f or or decisively
against any other purely semantical view f or that matter). The reason is
simple enough - each view, even those which (unlike Realism) wish to
tamper with theoretical terms, can surely set down a consistent syntax
and semantics for their version of science (i.e. specify a language structure
and define reference, denotation, satisfaction, truth in the usual model
theoretic fashion), and in the cases where the semantics prevents exact
reconstruction of the customary language of science simply deny that the
SYSTEMATI C REALI SM 411
residue belongs to science proper. I do not see how to refute any such
normative position in itself, since nothing is argued and consistency is
granted. No, what occurs in the literature are proposals for semantics, and
arguments against proposed semantics, based upon other considerations,
e.g. upon epistemological desiderata. Realism as such has never been ar-
gued against directly, only indirectly through the world view in which it
is embedded (and so long as the alternatives to Realism remain norma-
tively hard-nosed about the definition of science, the same is true for
them). Moreover, this procedure has been fair enough, in my view, be-
cause of the intimate way in which semantical theses are in fact linked to
the remainder of a world view.
This then is my first revolution in philosophy of science (it is a 'little'
one): t o recognize explicitly the larger context in which Realism is eval-
uated. I shall hope thus to escape from the misleading and confusing
nature of much writing on the subject which leaves these connections
only tacit. I shall be aiming to explore a Realist world view bot h to see
which combinations of doctrine are possible and which is the most plaus-
ible among them. But one can only offer up that latter doctrine for the
acceptance of the rational man - precisely because it and its non-Realist
alternatives constitute esentially entire world views I can see no way t o
further argue the issues. To those who value argued resolution above all
else there remains the challenge: find a neutral meta-language in which
the issues between the alternatives don' t reduce to value judgments among
components of world views (e.g. preferred epistemologies) that cannot be
further argued for (assuming each set of value judgments self-consistent).
Incidentally, i f I am right about this then it constitutes a strong argument
for what Feyerabend said long ago and everyone ignored (or failed to
grasp): the choice of an epistemology (enlarge this to world view) is in
some sense an ethical choice, (see [63]). 3
A second, larger revolution in philosophy of science is that now being
brought about by Feyerabend, Hanson, Kuhn, Lakatos, Toulmin, and
their ilk (el. [29], [30], [31], [32], [33], [44], [45], [81], [82], [131], [132]),
and by many other factors, cf. my comments on the third revolution
below. Traditional philosophy of science regards science as an abstract
structure, specified in abstract linguistic terms. As such it is entirely sep-
arated from any activity of the scientist qua human being; one does not
consider the psychology of discovery relevant, only the ' problem of in-
4 1 2 c . A . H O O K E R
duction' , one does not consider the decision process for testing theories
and choosing among them (which involves all kinds of value judgments,
as well as judgments of other kinds), one considers only ' objective' criteria
of acceptance and rejection. The consequence has been ' rational recon-
structions' of the history of science which entirely ignore large segments
of it (discovery is irrational) and in which scientists habitually contradict
the alleged canons of Science (e.g. by retaining a refuted theory). Science
is not a human discipline, except accidentally. When we press this view
on the issue of fostering insight into the course of science on this planet
we discover a tacit ideology lying beneath it: The science of science should
also be Science. But this leads to the exclusion of all that is importantly
human from the account, since most of the human activity cannot be
captured in the combination of basic descriptive statement plus logical
machinery. Indeed, the approach fails on its own terms since, even be-
haviouristically construed, specific scientific theories (e.g. of psychology)
are needed to give an account of how science is possible for human beings,
thus completing an epistemological circle in what was designed to be a
rigorously logically (hence epistemologically) linear space, the space of
scientific ascent to knowledge.
As against this abstractive approach the new revolution proposes placing
(or re-placing) the scientist at the centre of the social-intellectual-ethical
complex known as science. It proposes first studying the scientist, his
knowledge and values, principles of rational action, ideological and social
position, metaphysical beliefs etc. in order t o understand him (or her!),
then once a rational account of the lives of scientists has been offered we
can turn t o the task of asking in what measure an abstract notion of
Science can be rescued from the collective participations of scientists. The
detailed, systematic development of this approach lies in the future, what
Feyerabend and Kuhn have done has been to offer us glimpses of the
great power of this re-orientation where e.g. what was irrational vis-a-vis
Science (even though it affected its course profoundly) can now be seen
t o be rational vis-a-vis the scientist, where the profound links between
the psychology of perception and linguistic usage in reporting can now
be properly integrated with the logic of theory support and where the
intimate interaction between scientific decisions, scientist's values and
social and cultural conditions now receives a natural place with the proper
importance. 3
SYSTEMATI C REALI SM 413
This essay is written with this new approach explicitly in mind. This
is not to say that I explicitly reconstruct a Science from an account of
scientists, for this concerns mainly methodology and the application of
epistemology t o the logic of acceptance. But I do attempt to construct a
Realism which is ready to be grafted on to this account of science. In
particular, the intimate role which the scientific account of perception
plays in the establishment of an epistemology and semantics, indeed the
way in which the theory of rationality and epistemology are made to
learn from psychology, are designed to make way for a ' humans at the
centre' approach. Elsewhere I have been concerned with the application
of this view t o the future of scientists (see [73]).
It is precisely the shift to a naturalistic epistemology and theory of
rationality that constitutes the third revolution. To repeat: it is an in-
disputable fact that over the course of the development of science our
philosophical theories have also been changing in response to scientific
discovery. One only has to l ook at the philosophy of Bacon, Descartes and
Locke e.g. t o realize that what was at stake there was nothing less than
the intellectual rationalization of the triumph of reason in science over
faith. The foundations for the future course of Empiricism, and later of
Conventionalism, were laid here as part of the programmatic attempt to
comprehend and integrate the ' new' way to knowledge and the picture
of man that was emerging from it, j ust as Instrumentalism had earlier
emerged from the debate in astronomy. And so it has been ever since.
Recent research in perception has taught us much concerning the extent
of the ' interpretation' the mind introduces, it has rendered ludicrous the
notion of pre-occurring conceptualized data; recent advances in physics
have destroyed the synthetic a priori of Kantian space (cf. [58], [110]).
The advent of quantum theory has cast doubt on the nature and appli-
cability of classical logic (cf. the status of geometry vis-a-vis relativity
theory). Advances in neurophysiology and psychiatry, and even purely
formal advances (e.g. in decision theory) have modified our concept of
rational behaviour. And so one could continue.
And yet ironically Empiricism and Conventionalism themselves are not
able t o cope with this feedback from science to philosophy. For they assume
a rigid intellectual hierarchy in which philosophical doctrine stands above
and before empirical discovery and cannot be touched by it, rather is it
the a priori touchstone for understanding and evaluating empirical dis-
414 C.A. HOOKER
covery. Thus logic, induction (read confirmation theory), semantics etc.
were all assumed given apri ori in philosophy. A fortiori epistemology and
theory of language and rationality were assumed provided a priori in
philosophy. It is not therefore possible for philosophy to learn from sci-
ence. This rigid hierarchy had one great advantage, no question of episte-
mological circularity arose. It has two disadvantages, it did not match
experience and it gave no plausible answer to the question of the origin
of a priori knowledge.
But there is abroad now another, contrasting conception of the situation
which takes seriously the contributions of those sciences which apply to
human beings. We might call this the ' new epistemology' and ' new ra-
tionality'. 4 The central contention of bot h doctrines is that the philosoph-
ical theories involved here are in no better position than our best scientific
theories of the relevant human capacities. Rational minds are well-func-
tioning minds where the latter term belongs to the (ultimately) unified
science of neuro-psycho-physiology. Similarly what human beings can
know and how they come to know it are questions belonging to the same
theory.
According to philosophical tradition after Descartes, i.e. when the
'epistemological turn' began, epistemology was prior to science. Un-
happily, the progress of science revealed that such epistemologies were
almost invariably based on poor armchair psychology - and no wonder,
given our self-ignorance during most of the period. In similar fashion,
philosophy made the attempt to dictate what is rational in advance of
understanding how the mind works.
It will be well to also bear in mind the changes which this shift brings
with it. First of all there arises the necessity for a clear account of the
dynamic interaction between normative and descriptive systems. Where
epistemology and rationality belonged to first philosophy this was un-
necessary since the relationship was strictly asymmetrical - First Philo-
sophy was a critic of science; but now that the two interact we must in-
quire how it is that human beings may alter their principles of action
rationally, even their criteria for rational action, in the light of experience.
This great task lies outside the scope of this paper, but I want to point
out that this forms here a second bridge between epistemology (and ra-
tionality) and ethics, for in each case we have normative systems in
dynamic interaction with experience. Likewise in the spirit of unity,
SYSTEMATI C REALI SM 415
epistemology and the theory of rationality come much closer together.
In the traditional philosophy it was supposed t o be rational t o reason in
the fashion epistemology dictated, to be sure, but the effective integration
of the two notions was small indeed. Now we have an entirely new setting
for these issues: a man will have a variety of goals and various utilities
attached to them and to the consequences of the various means to achieve
them, these will include epistemic goals; what it is rational for him to do
will be determined on the one side by the general nature of human beings,
their capacities etc. and the circumstances in which he finds himself and
on the other by some decision principles, such as the maximization of
utilities. This model will hold for actions to realize epistemic goals (e.g.
creation of a scientific theory) as for others. Which epistemic goals it is
rational to choose will of course in its turn be a function of our nature
and circumstances on the one side and critical reflection on the process of
acquiring knowledge (in the light of historical experience and theory) on
the other. In this context many of the traditional concerns of philosophy
are radically re-cast. Perhaps the most important change is that in our
approach to observation. Whereas traditional philosophy placed great
emphasis on the a p r i o r i analysis of sensation because of its epistemolog-
ical role, this pressure is entirely absent here and with it goes any special
concern over sensory atomism versus gestalt wholeness, or pure sensa-
tion vs. interpretation by the mind - rather we are free to accept the
findings of psychology as to what to count in any circumstances as ob-
servable. Correlatively, the notion of an observation sentence and its re-
lation to theory is freed, to await informing by psychology and history;
in principle all positions from a factual empiricist-style permanence t o
Feyerabendian pragmatic adjustment are open (cf. below). This in turn
opens up a range of possible semantical positions concerning theory, de-
pending upon the deliverances of psychology concerning human thought
processes and the nature (if any) of meaning. 5
In future, any interesting epistemology and theory of rationality for
science will have to take the feedback from science into account. What
we have is the logical form of a circle from an empiricist point of view;
actually, it is a spiral extending along the historical time axis: episte-
mological theory evaluates scientific method, scientific discovery informs
epistemological theory. A similar spiral operates for our conception of
rationality. As I pointed out earlier, one of the consequences of adopting
416 c. A. HOOKER
this line is the rejection of the old empiricist dualism between abstract
philosophic criteria of justification for scientific theories and psychological
accounts of individual scientists devising and defending those theories.
The present view dovetails with the development of a naturalistic episte-
mology and theory of rationality that would unify the account of science.6
Another consequence is that the unit of empirical significance must now
be considered as not j ust the whole of science and metaphysics, but the
whole of philosophy and science. Given that our world view is at all cohe-
rent, and apriofism rejected, we might have expected these consequences.
The tools for forging the new theory of the scientific enterprise are,
however, already being formed; we are developing decision theory, gen-
eral systems analysis, policy science and related areas, and the psychology
of perception and correlative theories of linguistic usage are now in-
creasingly well understood - for more comment and reference material
see my [73], cf. [140]. I digress here long enough t o remark on the partic-
ular application of this new orientation to another old problem in the
formal philosophy of science, the problem of induction. The inductive
enterprise seems t o have evolved through three stages. In the first stage
a quasi-logical machinery was sought such that, upon being fed ' the facts'
at one end, it would generate the true theory of them at the other (or, in
a later modification, the truest theory of them). In the second stage this
goal was given up as impossible and what was substituted was the goal of
specifying a calculus of confirmation of some sort that would attach a
degree of support or confirmation of some sort to theories. In some ver-
sions (e.g. Carnap' s) acceptability meant probability of truth, in others
(e.g. Popper' s) not; in all quantitative versions I happen to have l ooked at,
the inductive calculus was essentially a set of probabilistic inference rules
of some kind. The more sceptical inductivists of this stage 2 mold de-
cided that the most that could be achieved by this means was a compara-
tive assessment of relative acceptability between theories rather than an
absolute acceptability. Stage 3 appears when this approach is set aside as
t oo specialized and the problem re-formulated as a decision problem for
scientists rather than a formal problem for Science. This issue now be-
comes "Gi ven my present circumstances and utilities, which theories
should I accept (i) for belief and practical reliance, (ii) for testing next?".
Here one assumes that epistemic circumstances and utilities will play a
key role. (Notice that the two decisions called for may concern very dif-
SYSTEMATIC REALISM 417
ferent t heori es- a point often overlooked in the presentation of methodol-
ogies !) In this new context formal methods of assessment constructed for
the earlier problem may again play specialized roles as the occasion
demands. Here too, one can obtain a more unified approach t o the vari-
ous so-called paradoxes of support (paradoxes of confirmation, inductive
inconsistencies, lottery paradox, Goodman' s new riddle of induction) by
seeing them as all placing constraints upon the consistent formulation of
a theory of rational action, especially the sub-theory of rational accep-
tance. In what follows I shall say no more about the ' problem of induc-
tion' , the reader will understand that, whatever the final form of a theory
of rational decision-making we are able t o formulate, my Realism as-
sumes a stage 3 formulation of the issue. 7
This essay is written then against the backdrop of these three recent
revolutions in philosophy of science. In a sense, the most important part
of the essay has been written, the rest being detail. The reader is fore-
warned that this larger context at times affects the canons of acceptable
attack and defence considerably. (E.g. it is t oo easy, on occasion, t o mis-
take a legitimate rebuttal of framework for an attack ad hominem.) It is
my hope, though it is perhaps unlikely, that the specific Realist doctrine
t o follow, will do justice t o the breadth and unity of these three large
scale changes which promise so much.
2. THE EMPIRICIST ALTERNATIVE
Since historically Realists have been largely put on the defensive by the
domination of the empiricist tradition, it is appropriate t o digress at the
outset t o briefly characterize the empiricist approach, the better t o ap-
preciate the origins of the chief objections to Realism.
Historically, the central dogma of Positivist-Empiricism (P-Empiri-
cism) was the Lockean doctrine of ideas, and specifically that all ideas
originated in either sensation or reflection. Let us call this the first dogma.
This doctrine gave rise to the second dogma, that all claims can be ex-
clusively and exhaustively divided into empirical and analytic claims (their
origins, but more importantly, their justification, being grounded re-
spectively in sensation and reflection) and to the third dogma that all
empirical concepts (i.e. concepts putatively descriptive of the world) be
analyzable without remainder in terms of the class of simple, ostensively
418 c.A. HOOKER
defined concepts given directly in observation. These latter two doctrines,
which state the semantical core of P-Empiricism, then serve as the founda-
tion of the epistemological core of P-Empiricism, namely (fourth doc-
trine) that all empirical knowledge is grounded in indubitable, simple
observational claims and all analytic knowledge a p r i o r i . It is an imme-
diate consequence of these doctrines, especially of dogma 2, that inductive
support is a purely logical relation holding between evidence and hy-
pothesis, s This foundational epistemology is a (is the) most important
instance of the ' Doctrine of the Given' and P-Empiricism is completed
if we add the fifth dogma of Phenomenalism: the referent of the simplest
observational descriptions are the simple sensory ideas in the mind (or
sensations, or more recently, sense data).
Behind the dogmas lies a tacit psychology, the ' arena for sensations
plus logical machine only' view of mind and the correlative up-from-
tabula-rasa-via-sensory-imprinting-and-logical-manipulation theory of
concept and belief acquisition. Hume brought the view to its logical com-
pletion with his non-cognitivist theory of ethics and aesthetics, agnostic
theory of religion, dissolution of the person and inductive scepticism.
The power of this position lies in its presentation of a clear-cut seman-
tics and epistemology of science, but supremely in its avoidance of episte-
mological risk - for from the above doctrines it follows that, for every
empirical claim, the truth conditions for the claim coincide with the
epistemic conditions of acceptance for that claim, and bot h lie in the
realm of simple observation. Let us call this latter, powerful result the
Coincidence Theorem of P-Empiricism.
The dogmas of P-Empiricism are not mutually entailing, so it is open
to a person to adopt fewer than all; as a matter of fact there has been an
historical tendency to divide between dogmas 1, 3, 5 and dogmas 2, 4. 0
More recent Empiricism has dropped all but dogmas 2 and 3 and a weaker
version of 4 that follows if the analytic-synthetic distinction is retained but
either of the indubitability and/or aprioriness claims dropped (although
the latter is more likely to be defended than the former). A weaker version
still (W-Empiricism) replaces the ' be analyzable without remainder
in terms of ' in dogma 3 by ' attain whatever semantic content they
possess through their logical relations to the members of' . (There is
a further weaking of dogma 3 that would add to this last clause ' and/ or
through implicit definition by the theoretical postulates' , but to the extent
SYSTEMATI C REALI SM 419
that this notion is clear at all it endangers the empiricism of Empiricism.)
The beaut y of the full-blown doctrine is its comprehensiveness and unity,
the more one weakens the position the less philosophically interesting it
becomes, it becomes in fact a fragment in search of a 'world-view'.
Any of these versions of Empiricism provide sufficient grounds for in-
troducing the observational/theoretical distinction as fundamental t o the
analysis of science. They differ only in the conservatism with which they
treat the specifically theoretical part of theories: P-Empiricism and Em-
piricism reduce this part entirely t o the observational part, they support
in this context the 'Verifiability Criterion of Meaning' (dogma 3 restated);
W-Empiricism does not reduce away the theoretical part but leaves it as
an in-principle second class superstructure parasitic on the observational
part for sense and function - an inviting target for would-be Instrumen-
talists. (P-Empiricism and Empiricism essentially admit only one induc-
tive argument, that of generalization from instances, W-Empiricism has
a much richer, i f more problematic, inductive structure.) 10
All variants of Empiricism support the ' accretion' model of the history
of science, for the continued existence of educated humanity means the
continued accumulation of sensory experience and Empiricism holds the
assertions describing these to be established beyond theoretical criticism
(they are used t o criticize theories, not vice versa). Theories amount sim-
ply to increasingly effective organizations of this accumulating data and
to prediction machines for data to come.
The foregoing remarks suggest (but do no more than this) the way in
which the Empiricisms also find their place in a coherent world-view,
(one has a semantics and epistemology, but also a psychology, which
determine the basic outline, i f not the detail, of a detailed model of man' s
evolution and functioning), one opposed in very important respects to
that of Scientific Realism as I shall expound it. It is in this opposition
that the real bite of the dispute begins to be felt.
3. T HE CONVE NT I ONAL I S T ALTERNATI VE
Conventionalism is the view that the theories of science are systems of
conventions, without literal significance, whose use is j ust i fed by the way
in which they organize the dat a of science. Since it constitutes the other
major opposition to Realism I shall briefly outline the position.
420 c. A. HOOKER
Conventionalism then also introduces a fundamental theoretical/ob-
servational dichotomy; for this reason it has often drawn succour and
support from much the same doctrines as back P-Empiricism. But the
special impetus for the doctrine lies elsewhere for, unlike the Empiricisms,
Conventionalism does not attempt to reduce away part or all of the
theoretical hierarchy in terms of the observational level, it simply shifts
the status of the entire hierarchy without being especially interested in
altering its detail.
The root source of the conventionalist urge, like that of the Empiri-
cisms, is the epistemological agnosticism t o which Scientific Realism is
heir (see Section 4). If theories can neither be verified nor conclusively
refuted then are we not justified in regarding them only as convenient
fictions, elaboratefa¢ons de parler to organize data? And is not this the
only justified view? Moreover this point of view apparently obtains some
backing from experience in mathematics where one has mathematically
consistent systems each claiming to satisfactorily realize subject matter
(e.g. geometry) and which primarily differ only in some of the root defi-
nitions (e.g. that of straight line). This was, e.g. the route that led Poincar6
t o the position (see [137]).
Recently pro-Conventionalist thought has been given some impetus
by two consequences of a contemporarily fashionable Realism, that ex-
pounded by Feyerabend and others (e.g. Kuhn, Lakatos in many respects).
For this view claims that the semantic content of the observation language
stems from that of the theoretical language and not vice versa and that
theory determines semantic content in such a way that conflicting theories
are semantically ' incommensurable' . But this suggests that there is no
area of science where a simple matching of human thought and nature is
possible and that disputes between theories are not factual in the ' hard' ,
straightforward way Realism seems t o require. Theories then seem more
like initially arbitrary creations of the human mind which float free of
the world except insofar as we use them as conveniences to organize our
lives. Whatever this support amounts to it is not compelling (otherwise
we would have a reductio on this form of Realism), rather must it be
combined with certain philosophical principles stating a preferred world-
view (in' particular, a preferred epistemology) before it supports a shift
away from Real i s m- but for those who hold such preferences the belief
that this was the only acceptable form of Realism (a view I shall counter
SYSTEMATIC REALI SM 421
below), like the perception of the epistemic agnosticism of Realism, would
constitute a strong motive to l ook for an alternative world view (it hardly
follows that it must be the Conventionalist view).
It is clear enough that Conventionalism fits yet another world view,
one which would share some basic features with empiricist world views
but in which there would be a rather different account of the significance
of the evolution of the neurophysiological equipment of the human being,
the significance of language use and thought as well as in epistemology and
semantics. (E.g., presumably conventionalists would want t o give an
evolutionary account of speculative thought which makes it clear why
the operative mechanisms made plausible the conventional status of the-
oretical talk; j ust as Empiricism, when it is not Idealist, must offer an
account of the neurophysiological mechanisms backing the reduction
of every concept to certain simple, impressed concepts.) It is, however,
neither my responsibility nor my interest t o elaborate that view
here (and one does not readily find it done by conventionalists them-
selves).
Instrumentalism, the view that theories have no truth value status at
all but are merely convenient calculational recipes, I regard as a ' last
ditch' version of Conventionalism (cf. Section 4.4.) 11
These are the doctrines which form the systematic background for
attacks on Realism.
4. EPI STEMOLOGY ( I ) : GENERAL CONSI DERATI ONS
4.1. The Central Argument
The attack on the Realist comes in many forms but the underlying, basic
theme is always epistemological, the basic attack is always the same: "But
you cannot know - cannot k n o w- which of your theories is true, or even
which is near the truth, so what sense is there in saying that they are to
be taken literally?"
I unreservably concede the premise of this argument, that epistemolog-
ical concession follows from even a superficial characterization of the
logical structure of science, la
The general sense of this objection is that it is not reasonable to adopt
Realism in the face of this epistemological fact. The retort to be answered
422 C. A. HOOKER
is " Why not ? What is the argument t hat leads f r om epistemological un-
certainty t o rejection of semantical t heor y?".
4.2. A Strong Form of the Argument
The strongest case t hat could be made woul d be one t hat t ermi nat ed the
ar gument wi t h the conclusion t hat Realism was semantically void or in-
coherent, i. e. literally made no sense. To achieve a valid argument we
clearly require a premise t hat passes from epistemological status t o se-
mant i cal status. The weakest appropri at e premise would be one of the
general sort: those propositions whose t rut h values one can be confident
of not knowi ng for certain can also make no sense. I shall call this assump-
t i on the Epistemological Dogma. Of course stronger assumptions will
also suffice. One of the strongest is: those putative proposi t i ons whose
t r ut h condi t i ons we cannot be certain of knowi ng (verifying) also make
no sense. This I shall call the Security Dogma. Ot her positions fall be-
tween these extremes, the most common probabl y being: those putative
proposi t i ons instances of whose t r ut h condi t i ons. . , etc. as for the Security
Dogma - this I shall call the Verifiability Dogma.
There is onl y one ki nd of general doctrine I know of t hat is even capable
in principle of support i ng any of these claims and t hat is Empiricism in
some form. Empiricism t hen is a general presupposition of the plausibil-
ity of this general ki nd of at t ack on the Realist. Empiricism has, of course,
been under increasingly severe at t ack for several decades. What emerges
from a careful consi derat i on of t hat criticism, however, is t hat empiricism,
despite the heavy criticism (justified in my view), is at its best a compre-
hensive philosophical doctrine which can onl y be adequat el y answered
by anot her equally systematic (comprehensive) doctrine (otherwise the
argument s against Empiricism largely become question begging). Such
must Real i sm be - for it is sure t hat Realists must reject the dogmas of
Empiricism (except perhaps a weakened dogma 3) since t hey leave no
possibility in the account of the world for a serious role for the specifically
theoretical in its own right, at all. 13
Whereas Realism and Empiricism seem at least t o be able t o come t o
grips over specific arguments, Convent i onal i sm seems best represented as
simply offering an alternative view of the world which one does not argue
t o from Realist epistemology but which one might find attractive i f one
SYSTEMATI C REALI SM 423
did not relish some features of Realist epistemology. (Actually, Empiri-
cism could also be presented in this way.) What are we t o say of it7
The first thing to be said is that Conventionalism shares many features
with Empiricism and to that extent is open to the same kinds of criticism.
Thus the conventionalist is committed to some form of the observational/
theoretical dichotomy (since primary empirical content is confined to a
class of privileged, observational terms - though there is perhaps a sec-
ondary empirical content expressed in the form "I t is a fact that the
conventions C allow an adequate construction of empirical knowledge")
and to a corresponding psychology of perception and reflection (the latter
with a conventionalist twist) to back the view. But this dichotomous ap-
proach has been severely criticized in its empiricist guise and the Realist
will judge the criticism no less relevant here.
Moreover, Conventionalism unsupported is, as Wisdom remarks, on
somewhat shaky ground for the following reason: it is not at all obvious
that a suitable set of conventions will be forthcoming that will secure
j ust those sentences a given theory requires. E.g. that the empirical dif-
ferences between Newtonian and relativistic physics might be recaptured
by suitable changes in geometric conventions concerning the notions of
time, straight line etc. is, i f it succeeds at all, a brilliant piece of theoretical
insight and not at all a superficial linguistic manoeuvre. To render every
scientific change as a change in conventions will undoubtedly require
considerable stretching of the notion of a convention. In the long run the
position must be forced to no more than this: to change conventions is
j ust to change the meanings of terms. This position is distinguishable
from Realism only if the meanings of terms are so chosen that the theo-
retical claims come out analytically true under the semantical conventions
(they become true ' by convention' ). The conventionalist is then faced with
three problems: (i) how to make it plausible that fundamental assertions
of actual theories are indeed analytically or conventionally true, (ii) how
to give a plausible account of the derivation of empirical generalizations
(at least of those which differ from theory to theory, i.e. the non-trivial
derivations) from such theories, (iii) how to give a plausible account of
theory change.
The instrumentalist, as I view him, steps into this situation and offers
to relieve the conventionalist of this burden by construing theories so as
to be devoid of semantical content (there is a metatheoretical language of
424 c. A. HOOKER
inferences t hat has semantical cont ent , but this is an entirely different
issue). 14 The instrumentalist t hen is the least commi t t al of all - he also
takes the intellectual life less seriously, even the t ri umphs of applied non-
euclidean geometry are written off as nought but a successful pragmat i c
switch in calculation technique.
Even so, this extreme move can certainly be made in order t o prot ect
the position. For t hat mat t er I have no doubt t hat a consistent conven-
tionalist t reat ment of science coul d be offered, sufficiently hard-nosed t o
mol d science t o suit it rat her t han vice-versa. Conventionalism, like Em-
piricism, can be t urned i nt o a sufficiently elaborate world-view and has
enough initially plausible cont act wi t h the world t o be defensible by suffi-
ciently hard-nosed techniques. The Realist t hen can onl y pit against this
his own systematic position. It behooves me therefore t o present a Realist
epistemology t o replace t hat of Empiricism.
4.3. A Realist Epistemology
Now, and I cannot stress this t oo heavily, whether or not a Realist gets
into difficulty here depends largely upon the epistemology he chooses. If,
e.g., he chooses t o adopt the Doct ri ne of the Gi ven (dogma 4), i.e. a
f oundat i onal epistemology, t hen he will be hard put to it not t o plausibly
choose the given f r om the domai n of the sensorially experienced; i f we
now add (i) t hat theoretical concepts are not i nst ant i at ed as such in this
domai n and (ii) t hat the theoretical description and the initial observa-
t i onal description (the ' scientific i mage' and the ' mani fest image' ) of the
sensorially accessible situations are in general in conflict (bot h plausible
claims) tl:vn we must conclude (twice over) t hat theories are relegated t o
a status where t hey can never funct i on as literal descriptions. Dogma 4,
therefore, had better not serve as part of any Realist epistemology.
Before we go any furt her it will be well t o state more exactly the con-
sequences of rejecting a foundat i onal epistemology, for such a rejeetion
may be, and usually is, associated wi t h a more general philosophical po-
sition of a sort considerably more ' r adi car t han the rejection actually
requires. Precisely, I want to poi nt out t hat the rejection of a f oundat i onal
epistemology does not entail (1) t hat one cannot draw a legitimate ob-
servational/theoretical di chot omy, and does not entail (2) t hat there are
no epistemological asymmetries among the claims of science, so t hat these
SYSTEMATI C REALI SM 425
claims can be ordered in respect of epistemological primacy or accepta-
bility (in particular along some axis which might correspond to an ob-
servational-theoretical hierarchy), and does not entail (3) that semantic
content is purely non-ostensive and that the notion of ostensive definition
must be given up. (The lee-way opened up here is important for a dis-
cussion of Realist alternatives.) Now we are ready to proceed.
There seems to be just one epistemology f or the Realist and that doctrine
is dictated by experience, namely a naturalistic, foundationless epistemology
of ignorance. By tiff s latter I mean (1) that epistemological theory is strongly
determined by our scientific view of the world and our place in it, 15 (2)
that with respect t o justified belief man commences from total ignorance
and approaches the truth only through imaginative guess-work and trial
and error, and (3) there is no class of beliefs having the privileged status
of unrevisable truths (in the historical process of critical inquiry all beliefs
may be subject to revision, however ' fundamental' ), there is therefore no
class of accepted truths that could serve as the foundation of empirical
knowledge.
The Realist is pledged to take his science seriously, in particular, there-
fore, the sciences of Biology, Anthropology, Psychology and the History
of Science itself. The former two sciences inform us of the evolution of
man from more restricted origins, the initial ignorance is well attested -
even information built into neurological structure through evolution in
the form of dispositions etc. can be called in question when it appears at
the conceptual level and anyway could not merit a role as foundation for
knowledge on those grounds. (Though there is an important role for such
pre-experiencial structuring - cf. my [66].) Psychological research in-
creasingly undermines the relevance, hence plausibility, of the pheno-
menaiistic approach to perception. In its place I insert my own Realist
view of perception, designed t o harmonise with the scientific findings;
for clearly i f the phenomenalist analysis of knowledge is accurate there
is no place for Scientific Realism - Realism in science includes psychology,
hence the theory of perception, and so requires a Realist philosophical
doctrine of perception also. le My Realist doctrine is unsympathetic t o
any ' Doctrine of the Given' , i.e. to a traditional foundational epistemology.
(For the detailed argument see my [58].) The history of science supports
the view that even the most fundamental assumptions (even the laws of
logic, let alone the fundamental ontology types, the nature of space and
426 c.A. HOOKER
time etc.) may be opened to challenging revisions. 17 Hence the doctrine.
So, man is born in ignorance and has only his pitiful senses and his
imagination as aids to truth. No theory can be known or be guaranteed
to be true - neither by observation nor by other means; the history of
science teaches us the overwhelming likelihood that today' s theory will
not be tomorrow's. Theories are only intended true descriptions. More-
over they are intended truths for which we must admit that there are
infinitely many incompatible alternatives, we must even admit that there
are infinitely many experimentally indistinguishable incompatible alter-
natives. And the Scientific Realist must admit that the reality which he
claims science intends to accurately describe is known only through these
attempts at theoretical description; there is no independent access. Thus
one must face each theory with a degree of critical or sceptical distrust- a
generalized agnosticism is the order of the day.
Let us face this bogey right here. According to the Realist, every time
one does serious science one sticks one's philosophical and scientific
neck out, one runs the risk of being wrong (without, I think, knowing
what the odds of error are). It is no use pointing to that conclusion ex-
pecting the Realist to capitulate in horror and stomach ulcers once it is
clear how wide the scope of his agnosticism must be. (Without reasonable
doubt, every decent extant scientific theory is false and pretty obviously
so.) For the Realist revels in this situation ! Remember Popper's dictum:
the more risks you run, the more interesting and imaginative your theory
is, the more you stand to learn through its failure.
On the other hand, in the face of these uncertainties many philosophers
turned away to look for more comfortable truths. Various shades of
Empiricism, of Instrumentalism, of Phenomenalism offer to take the un-
certainty out of epistemic life by reducing theories to harmless parasitic
non-entities of convenience, or by reducing them away altogether. (How-
ever clad in impressive theories of concept formation etc., this is the truest
function of the foregoing positions.) Scientific Realists, however, gladly
accept these uncertainties and their consequences - for they have a real-
istic picture of man, they know the poverty of his ignorance, the slight
riches of imagination and the limited reach of science. By comparison,
the alternatives succeed, ironically, only in stripping man down to a
meagre stature - too meagre even to provide a plausible account of his
activities. This much we have already seen in the brief comparison of
SYSTEMATI C REALI SM 427
empiricist dogmas wi t h the scientific view of man. And we shall see it
reinforced when we compare the empiricist structure of science itself wi t h
a more realistic version.
Accordi ng t o this Realist analysis the epistemological dogma is t oo
restrictive twice over. In the first place there is the assumpt i on t hat it is of
t he essence of informative, communi cabl e cont ent t hat it be possible t o
be certain of knowi ng it; this the Realist denies citing as the out st andi ng
count er example precisely the cont ent s of scientific theories, is In addi-
tion, there is a tacit assumpt i on t hat knowledge essentially demands cer-
t ai nt y and this the Realist also denies, for it represents an impossibly hi gh
goal. (On this criterion, even the logical form of the world woul d be ex-
cl uded f r om knowledge.) We may insist on this criterion of knowledge i f
we wish, but we shall t hen simply call fort h a new word t o refer t o all t hose
things of i mport ance t o us which are not ' known' . What the Realist may
assent t o (this Realist woul d anyway) is t hat knowledge does demand
simple t rut h, and t hen claims t o knowledge can still onl y be tentatively
accepted. For Realism, knowledge does not t urn out t o be a very im-
por t ant c a t e gor y- and sure or certain knowledge a completely insignif-
icant cat egory - the interesting cat egory is tentative knowledge and con-
jecture. This result may seem ironic, even paradoxical, t o many philos-
ophers who woul d regard Realists as over-bold about their claims for
man' s abilities (e.g. the ability t o transcend sensory experience), but
act ual l y it expresses a nat ural agnostic modest y which flows precisely
f r om t aki ng man' s ability t o theorize seriously.
A little earlier I poi nt ed out t hat the real power of P-Empi ri ci sm stems
f r om the fact t hat its reductionist principles collapse t r ut h condi t i ons
i nt o knowledge conditions, so t hat no epistemological agnosticism was
necessary. The Realist rejects such semantics. Here we set down the first
t wo semantical distinctions of Real i sm:
(D1) The t r ut h condi t i ons of a proposi t i on are distinct f r om the con-
ditions for its havi ng a coherent semantic cont ent (i.e. for its
meaningfulness).
(D2) The condi t i ons under which a proposi t i on is true (respectively,
false) are distinct from the condi t i ons under which it is known
t o be true (respectively, false).
These distinctions are theoretical distinctions; of course a single set of cir-
cumstances may in f a c t yield even all three sets of condi t i ons (e.g. the
428 C.A. HOOKER
circumstances of our world) but even here it is a rare thing for the minimal
suflieient conditions of each sort to have any signitieant overlap, let alone
coincide, t9 Now the Realist is faced with a task: specify the semantics of
scientific theories. I shall return to this task later in the essay.
4.4. A Weaker Form of the Argument
It is time now to return to the original argument and, having rejected the
stronger versions of it (where the valid conclusion was literal senselessness
of Realism), consider a weaker, but more plausible version. Now the
argument runs: since no theory can admittedly be known to be true,
since in fact all those that we have are almost certainly false, and since we
change theories continually, is it not prudent.., wise.., reasonable! that
we take a more conservative line than Realism over their significance?
(And this is just what Conventionalism-Instrumentalism does - it identi-
fies the semantical nature of theories with one of their useful non-epistemic
functions, as convenient calculation devices.)
The Realist ought to be quite candid about this point: there is nothing
in the realm of the ' hard data' of individual theories that could possibly
distinguish his position from the Instrumentalist alternative with respect
to those theories. From this highly restricted point of view the most that
can be done is to point to the central role which theoretical structure
plays in guiding research, modifying and changing scientific beliefs, form-
ing hard data from phenomenal data etc. The hard-nosed opponent, how-
ever, can simply treat this as heuristics.
But we can enlarge the battlefield. The point of Realism is to support
a naturalistic world view. In this world view objects, geometry, man, the
universe at large, are taken seriously as science knows them - and science
knows them basically in terms of its theories. The instrumentalist on the
other hand must restrict himself to organizing observable data, he can' t
take this view of the world seriously, the only things he can take seriously
are the little partial glimpses of it that come to man through his senses.
The final protest that the Realist makes then is to point out the implausi-
bility of the resulting anthropomorphism that takes only things at a man-
sized level seriously, the likely unhealthiness behind the insistence on
epistemologieal security, the in-built tendency towards solipsist pheno-
menalism and so on. But exactly the point of being hard-nosed is the
SYSTEMATI C REALI SM 429
willingness (I am t empt ed t o say, the ability) t o ignore these charges. And
so at last we r un up against differences of philosophical t emperament .
The conservative, the Inst rument al i st , is willing t o write off these reasons
for t aki ng theories seriously as literal descriptions as mere appearance
against the gai n (for him) of t aki ng no epistemological risks wi t h theories
- which are so manifestly risky - while still giving their skeletons some-
t hi ng t o do in science. The Realist on the other hand finds t hat the risks
onl y add interest t o a view al ready satisfying because of the way it comes
fairly t o grips wi t h the unfol di ng intellectual dr ama of the human race.
He woul d like an argument, some specific feature of science t hat could be
adduced as a reason for not t aki ng a st rai ght forward view.
There are onl y t wo addi t i onal argument s I can t hi nk of. The first runs
as follows. The simple schematic Empiricist model of science (Not e 12)
places no restrictions on what can go i nt o the t opmost layers, but wi t hout
restriction it woul d be possible t o trivially explain anything. In fact sci-
ence empl oys a host of selection principles which operate at this level t o
severely restrict the ki nd of t heory admissible - the most i mpor t ant of
these are simplicity in all its forms (often incompatible forms s0), retention
(in a generalized way - e.g. as a limiting approxi mat i on) of as much of the
preceding theories in a field as is possible, preferred kinds of development
t o suit preferred ontologies (e.g. t o suit at oms + void, or field, ontologies).
But these principles clearly have no i ndependent justification in nature,
t hey are empirically arbi t rary (t hough t hey are not pragmat i cal l y arbi-
trary) restrictions brought down t o make t he j ob of t heory const ruct i on
manageable. But i f theories in science are t o be governed by such obviously
ant hr opomor phi c rules how can they, shoul d they, be t aken seriously as
literal descriptions?
To this there is no reply but ' Why shoul d t hey not be?' It is t rue t hat
restricting our t heor y const ruct i on in these ways is reason for greater
scepticism about our theories, but why shoul d it be anyt hi ng more? More-
over, we still st and a chance of discovering whatever there is t o discover
in t he worl d at the level of the simple mat hemat i cal forms wi t h which we
operate. One mi ght as well argue: The onl y empirical i nformat i on avail-
able t o us comes via our pitifully limited senses and the i nst rument s we
can const ruct which are restricted by our pitifully limited resources and
i nformat i on; hence i t is useless t o t ake such ant hropomorphi cal l y ground-
ed theories seriously. Our i nformat i on is indeed severely limited but i f
430 c. A. HOOKER
anyt hi ng is necessary in a guide to the t rut h, our sensory and i nst rument al
experience is.
Thus this well-known move against Realism also fails. We have instead
anot her Realist distinction:
(D3) What is useful in a particular respect in science is distinct from
what is meani ngful in science.
Moreover, the premise of the preceding argument is not obviously true.
I am of the firm opi ni on t hat it is possible t o be critical of one' s preferred
metaphysics in science, as critical of it as we can be of anyt hi ng in sci-
ence. 21 Though the story is complex, the elements will be sketched a little
later. 22 Of course this means t hat , consistently wi t h the Realist episte-
mol ogy enunciated, there is ultimately no preferred metaphysics (just as,
in this sense, there is ultimately no preferred logic, or preferred anyt hi ng
else of this sor0. Moreover since I hol d the mat hemat i cal structures of
science t o embody a metaphysics, the cont i nual preservation and develop-
ment of a mat hemat i cal formal i sm is as critically justified as is the ex-
pl orat i on of the potentialities of a metaphysics. Finally, there are con-
siderations of the sort Popper offers 23 t o believe t hat there is some justifi-
cation for choosing t o explore the simplest theories first (t hough this
defense is easily in the worst shape of all). 24
There is of course another, well-known argument for not t aki ng the
st rai ght forward view. It is based on the ability to deal exclusively wi t h
the purely observational part of a t heory (with levels 3 and 4 only). The
met hod of capt uri ng j ust the observational part of a t heory is due t o
Craig. The argument t hen runs: Either the theoretical terms of a t heory
serve their purpose in leading t o the establishment of a net work of con-
nections among observable terms and t hen t hey are unnecessary for the
empirical purposes of science, or t hey fail t o serve this purpose in which
case t hey are t o be rejected as useless; in either case theoretical terms serve
no useful purpose in science (Hempel [51], pp. 49-50). But this possibility
poses no t hreat to the Realist, from the fact t hat such a separation of
t heory and observation is possible nothing follows about the semantical
status of the terms of the separated component s (whether theoretical or
observable) - one could obt ai n a conclusion of the sort opponent s of
Realism wish only i f an addi t i onal premise was added t o the effect t hat
what is not necessary t o the sense-confirmable port i on of science has no
semantic content, but this premise simply recapitulates the empiricist-
SYSTEMATIC REALISM 431
positivist dogmas and will be rejected by the Realist, precisely because of
the anthropomorphic, implausible world view to which they lead (el.
p. 428 above). It is important to understand this point: this attempt is not
rejected because it has a defect (that would constitute additional reason)
but because of the general world view necessary to give it force. ~5
These arguments both fail and they are the only arguments I know of.
I conclude that there is nothing but philosophical temperament left with
which to decide the issue - except a reasonable assessment of the pluses
and minuses of each side, and for a realistic view of the world I'll back
Realism. (The aim - the fundamental aim - of science is discovery and
understanding, I think, not pragmatic convenience of control, nor power,
nor security.)
4.5 A Second Objection
The next objection hinges precisely on the tendency for theories to change
radically. If they do so then the presumed fundamental ontology of the
wor d may change, not just particular putative entities may now be denied
existence, but entire putative categories of entities, indeed all putative
categories of things, may now be denied existence. (This latter would
happen if, e.g., physics in general switched between a field-plenum and
particle fundamental ontology.) Moreover, so the objection goes, the
semantical contents of such disparate theories will also be radically dif-
ferent - even where many of the same phoneme groups are employed.
Moreover, it may well be the case that theories of these disparate types
are experimentally indistinguishable from one another. But if this is the
case how can any theories be taken as literal descriptions when the kinds
of world the various possible theories describe differ so radically from
one another? Or again, doesn't the radical semantical shifts that can
occur show that theories are only implausibly viewed as intended descrip-
tions of a single real world and more plausibly regarded as heuristic
linguistic devices?
The short reply to both questions is: Why? Where is the valid argu-
ment that leads from the premises to the conclusion(s)? There is nothing
very strange, given our Realist epistemology, with changing our minds
radically about the nature of the world; and what semantical shifts there
are follow on from that change of mind, as we attempt to express it. To
make the reply more honey-reasonable, and a little longer, the Realist
432 ¢. A. HOOKER
can point out that one can have many kinds of mutually radically disparate
maps of the one stretch of land - and language is like a complex map.
(It is the mapping view of descriptive language which explains why even
at our simplistic level of concept formation and mathematical skill we
might still expect to discover some truths concerning the world.)
5. E P I S T E M O L O G Y ( I I ) : T HE S T R U C T U R E OF S C I E N C E
Our considerations concerning the structure of science begin with yet
another objection. The first premise of the objection is that many, indeed
all, of the mathematically sophisticated mathematical theories of science
deal only in mathematical idealization (point masses, instantaneous rates
of change etc.). The second premise is that no such idealization is ever
realized in nature and the conclusion is that no Realist interpretation of
these theories succeeds. A weaker, but still sufficient claim would be that
not all such idealizations to which the theory was prima facie committed
can be realized in nature, and the conclusion retained in its present form. 2e
Certainly the Realist concedes at least the weaker second premise and
corresponding conclusion - with the important proviso that the word
'naive' qualify 'Realist' in the conclusion. But this is an objection of the
sort the Realist cherishes for it permits a complete turning of the tables
on the objector.
Let us begin by noticing that the empiricist is equally at a loss to cope
with this feature of science. Since we never observe such idealizations in
nature they must be theoretical entities, but then the theories in question
are patently ' about' such entities, rather than about the observable world
which is their only legitimate subject matter according to Empiricism.
The simple empiricist structure outlined in Note 12 simply cannot accom-
modate such entities, they remain anomalous. 27 (On this score, only the
conventionalist has a superficially plausible account.) But it is when we
begin searching for a Realist reply to this objection, which takes the form
of a more adequate Realist reconstruction of the structure of science, that
we see that both Empiricism and Conventionalism are bound to offer an
inadequate view of science and to ignore the profound complexities re-
vealed therein.
Elsewhere I have offered a preliminary investigation of what I termed
the 'global theories' of science (cf. [71]). These are theories that are inter-
SYSTEMATI C REALI SM 433
nally and externally global. By an internally global theory I mean a theory
that not only offers a fundamental model of the world, but also determines,
at least in part, what in real systems is observable, what instrumental
means can be used t o observe and how reliably, how strongly various
data support or undermine its assertions, even in what language the
dat a are t o be described, and so on. By an externally global theory I mean
one that requires a battery of additional theories for its successful applica-
tion. I argued that most of the theories of quantitative science were global
theories in bot h these senses. 2s
Empiricism, precisely because it treats theories in isolation from one
another, and because of its necessarily simplistic epistemology for science,
cannot accommodate these basic characteristics of scientific theories. (A
striking example of this is found in the fact that P-Empiricism and Em-
piricism, precisely because they reduce all theoretical assertions to logical
combinations of indubitable sensory reports, cannot erect a truly critical
theory of perception and hence cannot give a realistic account of psycho-
logy, for it is the business of a global psychological theory bot h to explain
how man behaves and how he comes to know what he does about what
he does.) For the structural schema for science which I suggested as a
replacement for the empiricist scheme see Figure 1. This schema has at
least seven respects in which it differs from, and improves upon, the em-
piricist schema. 29
(1) The external global characteristics are explicitly recognized in the
structure. The schema recognizes explicitly that two sets of additional
theories are in general an essential part of any scientific situation. Such
theories appear at 2 levels: (i s) there may be additional theories needed
in order to apply a given theory to a certain k i n d of situation (e.g. classical
and quantum mechanics of individual interactions must be supplemented
by classical and quantum statistical theory respectively before they are
adequate to handle practical applications to gases, liquids and solids); (ib)
having obtained the right general kind of theory, additional theories still
will often be needed in order to work out a theory of the specific k i n d of
experimental situation envisaged (e.g. to the quantum theory of pair for-
marion must be added the quantum theory of gaseous ionization and elec-
tromagnetic theory in order t o get a theory of the behaviour of such fun-
damental particle pairs in the general experimental environment of a gas
immersed in an electromagnetic field); ( i c) f i nal l y, in order to know how
434 c.A. HOOKER
to process instrumental outputs the theories of these instruments must be
set down and related to the various features of the theoretical situations
which may arise. These theories therefore have direct inputs into which
instruments are chosen (sometimes indeed even an influence on which
system (type) is chosen for investigation) and how the data is prepared
and processed (including, in particular, the estimates of its reliability).
In addition the structure includes the general 'background' theories - the
various theories of mathematics and psychology, and whatever else may
need to be assumed to obtain the ceteris paribus clause always invoked. 3°
(2) The internal global characteristics are explicitly recognized in the
influence that the theory nominally under test (To) has on these same
features of the situation, the selection of system and instruments the pre-
paration and processing of data. (I say' nominally' under test because there
is no severe asymmetry of roles between To and the other theories in these
respects.)
(3) The structure explicitly recognizes the important role of mathe-
matical models in global theories. The mathematical theories of science
that I have been discussing all model the real world using various mathe-
matical models (e.g. differential geometry, point masses, instantaneous
velocities etc.) and often these are known to be idealizations (e.g. point
masses). From a formal point of view the theory is directly about the
behaviour of these model systems, not the real world systems. Indirectly,
though, these models are seen as models of the real world systems, ade-
quate in the relevant respects in certain circumstances, within a given
tolerance. In most cases the transformation from the actual crudely ob-
servable properties of actual systems to the corresponding properties of
the model is a very complex one, requiring various 'corrections' and 'ad-
justments' to the crudely described system until the model system prop-
erties are approached. 31 And the further transformation from the real
system behaviour now described in the technical terms of the model to
that of the model behaviour may be just as complex, since real systems,
however described, do not behave exactly as idealized models. 32 The dis-
tinction between these two kinds of transition is important in establishing
the character of two distinct kinds of correspondence rules - of. (5) below.
I emphasize that it is not description in common sense terms alone
that requires the complex transition to theory discussed above. Even de-
scriptions couched in such technical terms as 'Wilson cloud chamber'
SYSTEMATI C REALI SM 435
'ionizing particle trajectory' , 'flip-flop pulse out-put' and so on still require
the same kinds of transformations before the relevant theory of these
systems can be brought t o bear.
I t is only when data are presented at the level of, and in the f orm of, the
theoretical model that we can obtain an experimental verdict on the theory
(or, i f you prefer, it is only when the theoretical model can be linked to
the simplest level of observational description via the theory of the ex-
perimental situation, that the theory can be confronted by experiment).
Thus the appearance of the ' model-system filter-transformer' in the
schematic diagram. All connections between theory and experiment pass
through this filter-transformer which has the task of commencing with
an identification of the objects in the experimental situation (system under
test plus instruments) - whether in common sense or technical terms makes
no essential difference when the goal is hard dat a (it only adds or sub-
tracts a step or two) - and tailoring their behaviour to fit the best the-
oretical models of them. Clearly it is this complex procedure which deter-
mines j ust how we prepare and process the dat a (as well as having the
prior function of influencing the choice of system and instruments). As
Sellars remarks, the empiricist reconstruction of science, however, sees no
essential function for models at all (see [117], p. 178).
With models taken seriously the deductive theory of explanation (of.
Hempel [51]) receives a necessary enrichment. The point of explaining the
behaviour of a system S is not j ust the deduction of true descriptions of
S' s behaviour from the theory - this almost never happens exactly and it
is far t oo crude a characterization of the actual situation. A theory ex-
plains S' s behaviour by providing an appropriate model for S such that
good approximations to S' s behaviour are deducible from that model
description in conjunction with the theory and via the filter-transformer.
One then says, provided the model is of the right sort (i.e. not based on
a mere formal analogy etc.) that one has captured what S really is, in
essentials. It is this provision of insight into the real nature of S that we
normally count on in satisfying explanations, az
Models play a similarly important role in inter-theoretical explanation.
What allows a number of distinct theories to be brought t o bear on the
same problem is the fact that they share aspects (or fragments) of a single
model in common. Moreover, one does not in general deduce the em-
pirical generalizations of a succeeded theory from its successor, this is
436 c. A. HOOKER
generally impossible because the two conflict, but the successor theory, by
providing a model in its own terms of the world as modelled in the suc-
ceeded theory, explains why the succeeded theory should have been as
experimentally successful as it was.
(4) There is a tittle observed feature of the scientific reasoning process,
namely that often not even the derivation o f observable results is deductively
valid. Traditional Empiricism assumes a fully deductive structure. The
situation as envisaged here, however, is to explicitly allow the theory/data
transitions appearing therein to include non-deductively valid arguments
as well as deductive arguments, s4
(5) The nature and role of the so-called correspondence rules is radical-
ly different and much more complex than that envisaged in the empiricist
model. Here there are two distinctively different kinds of linkages between
theory and the world.
The first of these are linkages between theory and hard data. There
are the connections usually considered by working scientists. In this case
the linkages are extremely complex, since they must properly be traced
down through the filter-transformer and back up through the data pro-
cessing and preparation chains. Notice, though, that the end result of
this chain can be expressed as a direct theory-hard dat a linkage. (E.g. for
quant um theory we might have, very roughly: "An electron-ionized track
intersecting a positron-ionized track in a Wilson cloud chamber gives off
a particle producing no ionization until.., etc.".) The temptation to resort
to this expression alone is fatal, however, because it suppresses all of the
complex and novel features of these linkages. (i) It tends to suppress the
distinction between hard dat a and phenomenal data (cf. below). (ii) It
suppresses the role o f models in developing the correspondences. (iii)
Even more importantly it suppresses their character as multiply theory
dependent. (I daresay there is not a single correspondence rule connecting
any real life global theory to hard data which is not thus dependent upon
other theories as well.)
But there is also a second kind of linkage included, that between model
and the phenomenal data and identification level. These are the linkages
that connect model theoretic assertions to commonsense descriptions, a5
(Thus: "A Wilson cloud chamber (technical model theoretic name) is a
vacuum sealed box capable of providing super-saturated vapour by. . . ",
and even further than this quasitechnical description: "A Wilson cloud
SYSTEMATI C REALI SM 437
chamber (same use) is a black metal box with a little plunger t hat . . . ",
and perhaps we are still not sufficiently ' far down' into commonsense
terminology. At all events we are dealing with a continuum of cases here
whose bot t om is ' pure ordinary English', whatever that is - cf. below.)
The scientist habitually makes use of such connections to tell his tech-
nicians, or the storeman, what t o do, but they never enter his account of
science. They are, on the other hand, what the epistemologically self-
conscious philosopher typically has in mind, though they are often con-
fused with the former rules.
Finally note that bot h types of linkage may contain inductive as well
as deductive components, again a departure from the traditional model.
(6) The present schema explicitly recognizes the presence and impor-
tance of a ' supra-theoretic' level of proto-theories and beyond those again
a further level of metaphysics. These elements, which are almost without
exception only tacit in the exposition of theories, are also almost univer-
sally ignored by philosophers. Positivist-empiricists were an exception,
they actually dismissed the crucially important metaphysical level! It is
hard to over-emphasize the importance of this level. Suffice it to say that
I believe that the entire history of physics can, in effect, be seen as the
competition of two incompatible metaphysical schemes - the atoms +
+ void and the plena + no void schemes. And I have recently argued in
some detail the thesis that the deep mathematical structures employed by
theories of mathematical physics is chosen so as to reflect the preferred
metaphysics that the theory is expressing (and further that this approach
offers the only hope of unraveling the mystery of quantum theory). 3e
The metaphysical schemes lay down the most fundamental general prin-
ciples governing the conceptual structure of a coherent description of
that kind of world - it is no wonder therefore that they have an all-
pervasive influence on the form and content of the theories that express
them. Roughly speaking, they can be formulated as meta-theoretical
assertions which place constraints on the admissable predicates and syn-
tactical forms for theory. (Toulmin' s ideals of natural order are a part of
this meta-theoretical machi ner y- cf. [131]). It is really much more en-
lightening to recognize a theory as having a meta-theoretical dimension
(to put it paradoxically) and it has been a great loss t o philosophy of
science to ignore this dimension for so long.
Also of considerable importance are the ' prot o' theories (Bunge's term,
438 c. A. HOOKER
who first drew my at t ent i on t o t hem - see [15]). These are the theories
t hat express the assumed general structures of some of the quantities
entering the formul at i on of scientific theories, and of some of the met hods
empl oyed in their application. Thus we include here a t heory of the struc-
t ure of the time cont i nuum (a set, ordered by an anti-symmetrical, transi-
tive, reflexive relation etc.), space, abstract measurement t heory etc. (The
reader shoul d consult Bunge for details - but beware of reading t oo much
i nt o these proto-theories, cf. my [71].) Such theories could be assumed t o
be absorbed i nt o the expression of each specific t heory (as I suppose we
ought chari t abl y t o assume the empiricists t o do), but it certainly is not
so in t ypi cal expressions of a t heory and their special role has been largely
overlooked. Of course t hey serve as an initial filling out of a metaphysics,
hence their position in the schema.
(7) Fi nal l y I draw at t ent i on again t o the role accorded a ' phenomenal '
or common sense level in the structure of science. We have already seen
how t hat role mi ght be criticized, indeed how in principle it mi ght be
eliminated, and how it is in any case not the chief linkage which appears
in scientific accounts. In bot h of these ways it differs radically f r om the
Empiricist analysis. A little later on I shall briefly describe may own cur-
rent views on its ul t i mat e fate. 8~
Ret urni ng now t o the objection with which this section began, namel y
why it is t hat the idealizations occurring in a t heory do not render Realism
untenable, we see now how a realistic analysis of science copes with the
problem. Those espousing this argument emphasize t hat a t heory' s sen-
tences do not in general accurately mat ch the phenomenal system bc-
havi our (a few low-level generalizations aside) and of course this is some-
times accurate. But t o conclude from thence (as e.g. Suppe [126] does)
t hat the t heory is false of the world is t o move t oo fast. Fi rst we need t o
separate out those cases of conflict between theoretical and phenomenal
level, i f any, which are t o be resolved on the side of theory. Such conflicts
can arise in situations where there is a fundament al conflict between the
concept ual structure of the t heory and t hat of the phenomenal level
(anal ogy: statistical mechanics vs ' phenomenol ogi cal ' t hermodynami cs).
In these cases it is probabl y the phenomenal concepts which must be
given up as conceptualizing mere appearance (unless the t heory has noth-
ing else going for it), t o do otherwise woul d be t o offer t oo much respect
t o concepts at the phenomenal level. In most cases of st rai ght forward
SYSTEMATIC REALISM 439
conflict, I venture to suggest, there is no direct conflict between phenome-
nal and theoretical concepts (e.f. below), it is just that some beliefs for-
mulated at the phenomenal level must be rejected, as In all these cases the
Realist position is unscathed, the theoretical terms offer the fundamental
description of the world and it is the phenomenal level which must be
corrected to meet it.
That leaves us with only those cases in which there are theoretical
reasons for believing that the theoretical idealizations cannot be realized
at any level. In this last case there are again two subeases - that where
there are grounds for believing that a particular idealization logically
could not apply at any level and the remainder where we have grounds for
believing that as a matter of fact it does not apply to our world (e.g.
respectively lines of force, frietionless surfaces). In the latter case we can
say that a theoretical claim is counterfactually true of the world, i.e. it
says how the world would really behave if the appropriate conditions were
realized. Contrary to those espousing the anti-Realist argument (e.g Suppe
[126]) this circumstance is precisely compatible with Realism, it is in fact
to say how the world really is in certain essential respects (and that these
respects are merely complicated by additional conditions). If we wish to
show this explicitly we may reformulate all relevant sentences of the theory
so that they take the general form 'As the circumstances C approach CI
(the ideal circumstances) the behaviour of the system approaches B~ (the
theoretically specified behaviour)', where we assume the limiting process
to be quantitatively specified. This leaves us with only the logically in-
applicable idealizations. These are fewer in number than people suppose, a9
A Realist must insist on rigorously eliminating them from theory. I know
of no case where this cannot be relatively simply done. And there are the
strongest logical reasons for doing so (namely, there are logical grounds
for asserting their ontic inapplicability). In short they can be seen to be pa-
tently devices of heuristic or convenience that can be done without. Thus,
with one or two insubstantial modifications (clarifications) of theory, Real-
ism emerges from the argument not only unscathed but strengthenedA 0
6. SEMANTICAL CONSIDERATIONS
6.1. The Theory of Truth
The Realist is clearly committed to a Correspondence Theory of truth.
440 C.A. HOOKER
Whether this in itself brings with it any special semantics is not easy to
say. Certainly Tarski's semantic theory of truth will be adopted; but,
contrary to Tarski's own view, I do not believe that this is an explication
of a peculiarly correspondence theory of truth, the coherence theorist
might - and I believe ought - equally well take it over, for it forms a
general account of part of the semantical structure of any language rather
than an acccount of a specific theory of truth. While the correspondence
theory is clearly connected with Realism, the coherence theory is con-
nected with Conventionalism and Instrumentalism where the ordering of
data is emphasized.
6.2..4n Objection to the Theory of Truth
These remarks carry me to a correlative objective. Look about you, this
objection runs, when does the scientific community count an assertion as
true? Just when it coheres with all of the data available and with the best
theory available. And what is the strongest operative criterion in the
preferring of one theoretical description to another? Why, aside from
compatibility with the data, the inner coherence and comprehensiveness
of the theory. Surely then if any criterion of truth is relevant to science it
is the coherence criterion, so that one ought to adopt this view of truth
rather than the correspondence view. In any case the foregoing episte-
mological considerations have shown that for a Realist the correspon-
dence theory would be effectively inoperative - it has already been con-
ceded that we cannot choose our theories on the basis of their correspon-
dence with the world, but precisely on the coherence criteria outlined
above. Once you leave all claims open to revision the only criteria re-
maining are coherence criteria.
Once again the Realist must candidly admit that such coherence criteria
dominate in science, and for just the reasons given in the objection. But
these criteria are criteria of acceptance, not of truth. The foregoing ob-
jection simply confuses acceptance and truth and in so doing collapses.
For the Realist to accept a claim or theory is a pragmatic decision to act,
for the time being, as if the claim or theory were true; for the claim or
theory to be true is a non-pragmatic semantical state of affairs in which
the theory in some sense accurately mirrors the world. Thus we have the
fourth distinction important to Realism:
SYSTEMATI C REALI SM 441
(D4) The conditions for accepting a scientific claim or theory are dis-
tinct from the conditions for the truth of that claim or theory.
Again (D4) states a conceptual distinction; most of the time the two
sets of conditions will in fact also be distinct but various degrees of overlap
may occur.
6.3. Realist Semantics ( I ) : The Issues
But now the central semantical challenges to the Realist are unavoidable:
(1) Say in what the meanings of theoretical terms consist, i.e. explain the
basis of the semantical contents of theories; (2) Show how it is possible
for theories to mirror the world, i.e. show how this semantical role for
theories is possible.
6.4. Realist Semantics ( I I ) : From Whence Meani ngs?
With respect to the first challenge I must admit that I have no satisfying
theory to offer. Indeed, this seems to me to be one of the outstanding
trouble-spots for Realism. But since, as it also seems to me, no-one else
has anything more satisfying to say than what I shall say here, this is an
outstanding problem for the Realist, but not a peculiar embarrassment
for him.
As a Realist it seems to me at this point in time that I have available
a spectrum of choices for a doctrine of semantical content, from a radical
position, say Feyerabend's or Quine's, through more conservative doc-
trines (e.g. Sellars) to a conservative doctrine quite close to the empiricist
tradition (Maxwell). I shall briefly elaborate these doctrines and then
state my own view of the most plausible position in the spectrum.
6.4.1. Quasi-Empiricist Realism
I shall begin at the conservative end, for that seems to be the least likely
defensible Realist position. Essentially, Maxwell takes a Ramsey treat-
ment of theoretical terms 41 seriously. Roughly speaking, theoretical terms
are treated as if they were bound variables, thus a sentence of the (sim-
plest) form
(x) Tx ,
is 'transcribed' to
(3t ) (x) t x ,
442 c. A. HOOKER
and so on for more complex sentences. As a mat t er of fact, t hough not
of necessity, this t reat ment is not extended t o the observable terms of a
t heory and so this approach presupposes, and enhances, an observational/
theoretical di chot omy.
It is clear t hat this approach has the i nt ent i on of t reat i ng the observa-
t i onal and theoretical levels quite differently; roughl y the i nt ent i on is t o
grant observational terms a ' full status' , t hey are t o have connot at i ons (a
sense, intensional cont ent or what-have-you) as well as denot at i ons (ex-
tensions), whereas theoretical terms are merely predicates, we know-not -
of-what-content, which st and in certain extensional relations t o each
other and to the observational terms. 42 Maxwell regards the observational
terms as essentially achieving their connot at i ons t hrough ostensive defi-
ni t i on (he speaks in terms of Russell' s knowledge by acquaintance and
knowledge by description). Moreover, in this exposition of a theory, the
observational level has epistemological priority for purposes of accep-
t ance or rejection of the theory. Thus this view builds in much of the
character of the empiricist view of theories - the issue is whether it builds
in t oo much t o be acceptable to the Realist.
First let us be clear t hat a number of i mmedi at e criticisms of this ap-
proach do not succeed: (i) deductive systematization is preserved and (ii)
it can even be plausibly mai nt ai ned t hat inductive systematization can be
preserved 43 (in any case this latter objection is not decisive against some-
one who abandoned i nduct i on altogether in favour of the hypot het i co-
deductive approach - again the Realist, qua Realist, may be non-commi t t al
on this latter choice); (iii) t hat alternative models satisfy the Ramsey for-
mal i sm is irrelevant; (iv) a Maxwell Ramseyan is not commi t t ed t o the
doctrine of the given (to dogma 4 of Empiricism), for nowhere is it as-
serted t hat the observational level claims are not open t o revision and
not hi ng has been said about concept-percept f or mat i on; (v) finally not e
t hat this position is not commi t t ed t o the analytic-synthetic dogma, for
a sharp observational/theoretical di chot omy is neither a necessary, nor
sufficient, condi t i on for this dogma (the one is semantical, the other
epistemological), so t hat the way is still open here t o regard logic as em-
pirical i f one desires (and this Realist desires to).
In fact, Maxwell claims t hat the Ramsey approach does not eliminate
theoretical rol es, for t hey have extensional roles not reproducible f r om
the resources of the observational level. I f ont ol ogy then is, roughly, what
SYSTEMATI C REALI SM 443
we must quantify over to have our assertions come out true, then the
intended ontology of theories includes theoretical entities. Actually, I be-
lieve he is strictly able t o assert only that very probably the Ramsey ap-
proach does not eliminate theoretical entities, for it is j ust possible that the
objects satisfying the existential claims are non-theoretical objects, t hough
this is so unlikely t o be true that Maxwell' s position is worth pursuing. 44
Moreover, this approach can consistently be conjoined with a Real/st
doctrine of perception and with the denial of the given t o recapture the
essence of the Realist epistemology adumbrated earlier. That observa-
tional terms are, de f act o, semantically and epistemologically privileged
counts nothing against Realism so long as observational claims are
equally open to criticism in the light of experience and theoretical ad-
vance. (That the ' nothing' here is strictly warranted is controversial, cf.
my treatment of Sellars' argument below.)
Indeed, Maxwell has a persuasive argument for his position. The ar-
gument runs as follows: (i) Beyond some level of some reasonably direct
observational acquaintance we could really only know about the nature
of the physical world through its causal consequences for those features
of the world which arc included in the directly observable level, and (ii)
that the only such deeper-lying features which appear on this ' observable
surface' are the structural relations and causal relations among the un-
observable entities (simply because such structural and causal relations
can be mediated through causal or nomic connections whereas the in-
trinsic properties themselves cannot be mediated in this fashion); it
follows, therefore, that theoretical terms can significantly only stand for
such relations and hence that they will be essentially structural in form and
without qualitative content semantically. 45
And there is a correlative argument for the observational/theoretical
dichotomy which runs: Scientific theories are developed to explain and
predict events in our observable world; however, because of the poverty
of our senses such theories typically must employ unobservable entities,
i.e. theoretical entities, in providing powerful explanatory and predictive
systems; accordingly, the observational/theoretical dichotomy is an in-
tegral part of the scientific endeavour and arises quite naturally out of
our sensory limitations.
The premises of these arguments seems to me to be entirely plausible
and the arguments are surely valid. Together they constitute a powerful
A.AA. C.A. HOOKER
prima facie ease for this quasi-empiricist position (indeed, for the em-
piricist position itself).
I turn now to an important argument of Sellars (see especially [116]
and [117]). Sdlars argues that terms have 'first class semantical status'
only when they belong, or could belong, to the picturing language, i.e.
only when, in some sense, they literally mirror or picture the world; but
they can only have this semantical role if they occur, or could~occur, in
the observational vocabulary. Therefore, any analysis of science, e.g. Em-
piricism, which relegates theoretical terms to an in principle non-observa-
tional role necessarily prevents them from being taken Realistically. It
looks as if this argument will apply equally against Maxwell's position.
Leaving aside the first premise, which is not strictly necessary (though
I am inclined to accept Sellars' semantical account of what is peculiar
to the correspondence doctrine of truth anyway), the Realist ought to
grant the other premise, and the validity, of the argument. Indeed the
recognition that, to be taken seriously (Realistically), the theoretical terms
must, in principle, be capable of playing an observational role is one of
the greatest of the post-Positivist lessons learned, primarily through the
efforts of a mild epistemologieal conservative (Sellars) and an episte-
mological radical (Feyerabend). Roughly the idea is that, from a God' s-
eye perspective, the world must be exactly as (true) fundamental theory
says it is and so one can in principle conceive of a sentient with sufficiently
rich sensory equipment to observe all of these characteristics. (Notice
though that this doctrine does not require commitment to any empiricist
belief that observational terms have a special semantical content, it is
entirely independent of that doctrine.)
But the further conclusion that the argument applies equally against
Maxwell's position as against Empiricism is too hasty, on two grounds.
(1) The argument is valid if, but only if, the ' in principle' is construed as
'logically possible'. In particular, the argument does not support the con-
clusion that it must be physically possible for theoretical terms to play an
observational role, let alone that this role be actually realized. After all,
whether or not this latter occurs depends, at least in important part, on
the actual physical structure of our sensory relations with the world and
on the structure of our central nervous systems (the former governing
which information is accessbile to us and the latter the possible develop-
ments of our conceptual-perceptual response apparatus). It is consistent
SYSTEMATI C REALI SM 445
with Realism, therefore, that the de facto epistemological privileges which
some terms enjoy should, as a matter of fact, be unalterable, as long
as they are in principle replaceable by theoretical terms (and the cor-
responding claims open to criticism etc.). And something of j ust this
sort seems t o be Maxwell' s position. 4~ (2) The position as Maxwell states
it, though already an elaboration of Ramsey' s approach, can be elaborated
still further. E.g. it is open to the Realist to claim that at some point some
of the theoretical terms either (i) actually, or (ii) may in physical principle,
or (ii0 logically might enter the observational vocabulary, j ust as Sellars
claims they must. 4~ I shall label these positions, respectively, radical, mild
radical and conservative quasi-Empiricism.
To see how even the mild-radical and conservative quasi-empiricist
Realists can reply to this objection, note that they t oo are eommited t o
the view that it is the satisfying substituents of the bound variables that
are the theoretical entities, hence they will claim that there are genuine
theoretical terms, namely chosen constants which designate these sub-
stituents, it is j ust that these constants have only denotations and no
connotations. Of course scientists speak and feel as i f their theoretical
terms have connotations, but the quasi-empiricist has several moves he
can make here; whether these su~ce to construct a genuine connotation
(at least one of these certainly improves the Ramsey approach) or ex-
plain the situation away as ' merely psychological' is of lesser interest. 4s
A presupposition heretofore has been that terms receive a connotation
of the sort that counts only i f instances of them are observable. The former
of the foregoing alternatives would break with this assumption (and right-
ly so it seems to me, I can see nothing but old empiricist prejudices t o
recommend it). It has to be admitted, however, that i f this assumption is
allowed to stand then full-blown theoretical terms fade beyond our actual
reach t o the realm of mere (logical or physical) possibility. In this case
the empiricist bite of these positions would make itself felt. The hard-
nosed Realist can, however, either choose to grasp this thorn (it is not
quite a nettle) and continue with his exposition of theoretical advance and
change eschewing all use of full-blown theoretical terms, or, as I said,
simply deny the supposition and choose the modified Ramsey approach. 40
If, on the other hand, a more radical line is chosen and it is conceded
that theoretical terms may be created and can actually usurp the role of
the observable terms, then the last of this collection of problems disap-
446 C. A. HOOKER
pears. Here once again observation as the source of connotation may be
safely maintained. There remains, however, to give an account of exactly
what is involved - conceptually and psychologically - when the transition
happens.
The entire range of positions, however, must also face the following
problem: What is the relation between the existing observational con-
cepts and claims made in terms of them and the possible or actual re-
placing theoretical concepts and corresponding claims ? Many people have
argued, and there seems strong grounds for believing it, that the world of
simple observation and the theoretically described world are incompatible,
that the conjunction of all the ordinary descriptions acceded to by com-
monsense and all the theoretically adequate descriptions cannot be simul-
taneously true (e.g. Sellars - the 'manifest' and 'scientific' images clash;
Feyerabend - common language embodies a false theory). This conclu-
sion would leave the conservative and mild-radical stuck with an obser-
vation language for science which is admittedly conceptually inadequate
but which they cannot get rid of. However, if it became absolutely nec-
essary, this thorn too could be grasped. On the other hand a distinction
might be made between the adequacy of the concepts of the commonsense
conceptual framework themselves and the truth-or-falsity of the beliefs
formulated in terms of those concepts - especially if one continued to
accept the analytic/synthetic distinction. Now it is not obvious that in
some deeper sense the concepts of the common sense conceptual frame-
work are inadequate, i.e. quite incapable of being used to formulate true -
or at any rate experience-supported assertions, as opposed to many deeply
embedded commonsense beliefs being in fact false. (Of course the com-
monsense conceptual framework is inadequate to formulate the deeper
intended truths of science, but this is compatible with its possession of
adequate concepts for the formulation of shallower truths.) Moreover
the sense of' replacement' of commonsense by theoretical concepts would
now be a rather benign one, indeed 'addition' might often be the more
appropriate description. But if this latter alternative holds then the con-
flict (though not the difference) between the two conceptual frameworks
is removed (though also not the difference in beliefs) and the painfulness
of the clash evaporated. In any event this is a line of response at least
initially open to the quasi-empirlcist, for it does satisfy Sellars' demand
that theoretical terms should in principle take on an observational role
SYSTEMATI C REALI SM 447
(one which I shall explore further below). What all positions are forced
to admit is that, whether in practice or only in principle, the observational
language is to become dominated by theoretical predicates when it comes
to an objective description of the world.
At this point the further radical alternative of denying the observa-
tional language any special semantical privileges (by denying the existence
of observationally conferred connotations) seems increasingly attractive,
since by removing any distinction in status between the two sets of terms,
except for the pragmatic one of being the terms in which we habitually
couch observational reports, we (1) remove the sense of any special break-
through occurring when theoretical terms enter the observational level
(a man is simply re-educated and his choice of observational terms change)
and (2) open the way to saying that the dash, i f conceded, between the
two sets of terms is essentially j ust a clash between two trial-and-error
attempts to describe the world, i.e. only a generalized version of a clash
between two theories. In the further account now required of the origins
of observational and theoretical concepts the variant claims would run
roughly as follows: the conservative and mild-radical variants would
claim that of physical necessity, or in fact, the world conditioned in us
responses requiring j ust (terms for) the privileged extensions, while the
more radical variant would claim that our conditioned responses are
actually, and not j ust in principle, open to conceptual alteration.
It should be appreciated that denying any special semantical status
to observational terms (1) does not entail a denial of the analytic/syn-
thetic distinction (el. Not e 28) and (2) does not entail rejection of the
not i on of ostensive definition. Once again this is worth emphasizing be-
cause these rejections, although congenial to a radical semantical position
and often found together (e.g. in Quine) do not inevitably go together. 5°
There is here then a rich variety of quasi-empiricist doctrines compatible
with Realism. I have elaborated them at some length because their com-
patibility with Realism, and their richness, seems to be insufficiently ap-
preciated.
6.4.2. Anti-Empiricist Realism
Paul Feyerabend stands empiricism on its head. sl In the first place, as I
have noted, he regards observational terms as simply pragmatically deter-
448 C.A. HOOKER
mined by various condi t i oni ng processes (cf. [30]). But t hen one can ask
" Fr o m whence are these terms dr awn?" Feyerabend' s answer is: Fr om
theories - a (sufficiently general) t heory determines its own observational
language. Correlatively, every observation language becomes ' t heory
l aden' - even so-called nat ural languages are regarded as embodyi ng
primitive theories. Not onl y is the doctrine of the given denied, any
' nat ural ' observational level is also denied. The observational/theoretical
di chot omy disappears from semantics, and nearly f r om epistemology,
i nt o pragmatics, where it is of relatively little importance.
It follows t hat for Feyerabend theories must determine their own se-
mant i c cont ent , hence also t hat of their observational sublanguages.52 This
suggests the not i on o f i mpl i ci t def i ni t i on - an obscure and difficult concept,
t hough some sense can certainly be made of the idea t hat the syntactical
structure of theoretical postulates restricts the range of possible physical
models for the t heor y and in this sense contributes t o the relations among
the terms. Feyerabend hi msel f has proposed a syntactical (' structural' )
approach t o i nt ernal l y determined semantical cont ent (cf. [33]) t hough
not , I t hi nk, very successfully (cf. [2] and my [70]).
At t empt s t o state clearly a plausible and viable not i on of implicit def-
inition are fraught wi t h difficulty. One of these is the danger of collapsing
the not i on i nt o t hat of explicit definition. In the present context this woul d
amount t o reducing implicitly defined terms t o observation terms ~ l a
Empiricism. For Beth has proven an interesting t heorem t o the following
effect: under certain general assumptions a predicate is implicitly definable
i f and onl y i f it is explicitly definable. ~a It seems t o me t hat Beth' s t heorem
deals a serious blow t o any plausible not i on of full-blown implicit defini-
tion. Serious, but not in itself deadly. There are essentially t wo ways
ar ound the t heorem: reject the criterion of explicit definability as t oo
weak, or t hat of implicit definability as t oo st rong. . 4 pr opos of the latter,
one could go on t o t al k of par t i al l y implicitly defined predicates which
were not required t o satisfy so st rong a condi t i on as Beth' s and so avoid
his t heorem but, so far as I can see, this is t hen t o say no more t han the
range of models is partially restricted by the axioms. Conversely, one mi ght
insist t hat explicitly defined terms be expressed by somet hi ng stronger
t han an entailed material equivalence, but the possibility of this of course
depends upon the resources of the language, in most cases not hi ng stronger
will be available. This suggests t hat we need the resources of a meta-
SYSTEMATI C REALI SM
449
language, and indeed there is a strong tradition concerning the nature
of implicit definition that formulates its approach metalinguistically; iron-
ically, it leads to equally great difficulty of an opposite sort.
For in the commonest tradition the notion of internally determined
semantic content is in constant danger of reducing to zero the empirical
content of a theory by rendering each of its fundamental postulates as
an (implicit) tautology. It suggests that a theoretical concept is, in outline,
the concept of a set of entities standing in certain relations (specified by
the theory) to certain other sets; but if this is the analysis of every term
in an assertion (or even of only an appropriate sub-set of the terms) the
assertion must become effectively tautologically true. 54 Ironically, this
analysis offers a restricted semantical content to theoretical terms which
is virtually indistinguishable from that offered by the Ramseyan analysis
in the quasi-empiricist approach.
It begins to appear that an account of wherein 'implicit definition' con-
sists can only itself be explicitly given on pain of either reducing theories to
tautologies or of reducing all implicit definitions to explicit definitions. 55
6.4.3. A Middle Way
Feyerabend and SeUars agree that ultimately theory must be able to de-
termine the observational level, they agree that the conceptual framework
of common sense and theory clash, and they agree that in some cases the
displacement of common sense observational terms by theoretical terms is
actually happening (cf. Sellars [117], [118]). They disagree only about the
actual nature and role of natural language in science up to, and including,
the present. For Sellars the conceptual structure of natural language is,
roughly, constitutive of the objects of common sense, as well as providing
the resources for formulating additional theories concerning them - in
the sense in which a theory, in the usual sense of the term, has an ob-
servable domain, described independently of it (namely, in the natural
language) and which it is about, natural language is not a t heor y. 5s
Correlatively Sellars disagrees with Feyerabend' s claim that natural
language, more precisely the level of common sense observation terms,
can already be dispensed with, for Sellars argues that science has not yet
provided a sufficiently complete conceptual schema with which to replace
it. In the meantime the natural observation language acts in a de facto
450 c. A. HOOKER
epistemologically privileged role as the basis for theoretical acceptance
and rejection ([117], p. 186f.).
The disagreements here are not j ust quibbles on Sellars' part, a mere
mat t er of fine j udgment of historical timing, for Sellars has a different
answer from Feyerabend t o our semantical question "Whence comes the
semantic cont ent of t heori es?" Accordi ng t o Sellars, theoretical terms -
and here, as wi t h Feyerabend, t hey are full-blown terms, i.e. predicate
constants with connot at i on and denot at i on, or at any rate will eventually
be so (cf. below) - have two sources of semantical content. (1) There is
implicit definition construed as cont ri but i ng onl y via the reduct i on in the
range of possible physical models and (2) there is analogical transference
of meani ng f r om common sense concepts t o theoretical concepts, this
transference being medi at ed by the specific model or models of the t heory
(roughly, the theoretical concepts and the common sense concepts, t hough
distinct, are said t o share a range of higher order attributes in common). 57
It is this latter component which is all-important, for implicit definition
alone is not sufficient t o grant a full measure of meaningfulness t o predi-
cate constants, instead, qua implicitly defined, t hey are onl y would-be
constants ([117], p. 176), whereas "t he distinctive feature of the use of
models and analogies in t heory const ruct i on is t hat the conceptual frame-
work of the t heory is generated by specifying the analogies which are t o
obt ai n" ([I 17], p. 181). Component 2 grants fullness of meani ng t o the
would-be theoretical predicate constants.
It is the eventually-to-be-abandoned observational language of common
sense t hat is itself the external source for the generation of semantic con-
t ent t o the successor scientific world view. And "Onl y i f we recognize t hat
the anal ogy. . , of the model i s not t o be const rued in terms of the i dent i t y or
difference of first order attributes, can we appreciate how the use of models
in theoretical expl anat i on can generate genuinely new conceptual frame-
works and j ust i fy the claim t o have escaped from the myt h of the gi ven"
([117], pp. 183-4).
Notice t hat , apart from a more complex account of the relations be-
tween common sense observation and theory, Sellars' position is very
close t o t hat of the mild-radical quasi-empiricist.
What Sellars first owes us is an account of analogical transference of
semantic content. This is nowhere spelled out in the detail required t o
cope with an actual t heory of science, t hough it is clear f r om Sellars'
SYSTEMATI C REALI SM 451
writings t hat some f or m of model i ng is central t o the process and i nt ernal
logical relations also pl ay an i mpor t ant role.
What Sellars now owes us of course is an account of how the common
sense concept ual framework acquired its semantic content. Any at t empt
t o i nt roduce a succession of frameworks related t o one anot her as t heory
is t o common sense not onl y does not answer the question, but merely
postpones it, while at the same time forcing us back on Neandert hal con-
cepts (if any) as the basis of cont emporary semantics. Any at t empt t o
stop the regress at some language by insisting t hat meani ngs in it were
directly conferred in observational experience is in danger of returning us
t o the worst of the empiricist dogmas. Sellars of course is aware of this
and at t empt s t o provide a somewhat different account of the emergence
of the common sense concept ual scheme. This account is the famous st ory
of our Ryl ean ancestors who possessed onl y crude behavioural concepts
at first wi t hout any explicit concepts of semantic rules but who gradual l y
enriched their language t o achieve the full-blown conceptual scheme we
have t o-day (see [116], Ch. 5). Though the general idea of an evolving
concept ual scheme is surely correct, Sellars' account is still clouded by
the empiricist' s insistence t hat observation brings somet hi ng special t o
concepts, the mechanisms of the semantic ' boot st rap' by which evol ut i on
occurred are vague and do not seem t o avoid the pressure t oward a re-
gressive language as seeking a pure behavioural reductive base. ss Perhaps
i f Sellars had given us an account of the origins of the meanings of espe-
cially the general concepts of the common sense framework we shoul d
have a clearer underst andi ng of the matter.
6.4.4. The Problem of Semantical Determination
The advant age of quasi-Empiricism is its obvious bowi ng in the direction
of the observation framework of common sense t hat we all in fact use and
its consistency wi t h our i nt ui t i on t hat unobservables can onl y deal in
causally medi at ed structural features of the world whereas perceptibility
confers an addi t i onal qualitative cont ent t o concepts. Even Sellars accepts
these features. Its di sadvant age is t hat it seems to restrict theories t o a
much more poverty-stricken role t han we are inclined t o give them, for
whatever our doubt s about t hem, historical experience tells us t hat the-
oretical ideas have domi nat ed the course of scientific knowledge al most
452 c.A. ItOOgEg
everywhere and that, whatever their semantic defects, theories do seem
t o provide a ' picture' of the true nature of the world. Sellars accepts these
features too. The Feyerabendian position at the opposite extreme to
quasi-Empiricism has the advantage of taking theories seriously enough;
i f anything it has the disadvantage of taking them so seriously that they
cannot fulfill the role then demanded of them (and the role of an observa-
tional language is perhaps taken in t oo cavalier a fashion). The mid-way
position occupied by Sellars seems to retain the attractions of the extremes
but actually it also seems to possess transpositions of their respective
weaknesses. Theories are restricted, ultimately, t o logical combinations
of would-be predicates (the replacement of the common sense world has
yet to occur) depending for their content on analogical transpositions of
what are ultimately crudely observational concepts. (Yet i f theories are
t o be truly semantically transcendent the origin of the additional semantic
richness escapes us after all, or is itself reduced t o observation.)
We seem left then hanging somewhere between empiricist observa-
tionally conferred content and pure implicit definition, with perhaps a
third kind of source, as yet unspecified, thrown in (pure imagination
perhaps?), to account for the semantic content of theories. And this is,
essentially, where I shall be forced t o leave the issue. To become any
further engrossed in the detail of contemporary proposals for the analysis
of semantical content would take us altogether t oo far afield. The point
I wish t o stress again is the rich range of these positions which are com-
patible with Realism.
6.5. Realist Semantics (III): The Criteria of Realism
Our second challenge, it will be recalled, was to show how Realism was
possible, i.e. to show how that semantical role for theories was possible.
This challenge turns out to have no better answer than we can provide
to the first. For it is perfectly clear now what must be shown, namely j ust
how theories can come t o dominate the observational level - at least in
principle. But the mechanism offered for this process will clearly depend
in a general way upon the accounts offered of the origins of semantic
content for theories canvassed in the preceding. For Feyerabend observa-
tional languages are simply replaced tout court when theories replace one
another - the satisfaction of this criterion offers no special problem, it is
SYSTEMATI C REALI SM 453
assumed satisfied f r om the beginning. Sellars offers us his complex mech-
ani sm of analogical transference f r om the common sense framework,
but does not act ual l y say j ust how theoretical terms acquire a ' report i ng
role' from there, perhaps the t rai ni ng mechani sm is not unlike Feyer-
abend' s (see [27] and [24], pp. 211-218; cf. my [62]). The quasi-Empiricist
account as I have developed it has not hi ng t o say on the mat t er, not sur-
prisingly since for all but its radical extreme it is not a mat t er of act ual
occurrence; still it owes us an account of how the t ransi t i on mi ght go in
principle, this will however be of a somewhat different sort f r om the ear-
lier possibilities canvassed here since we woul d presumabl y be discussing
a count er-fact ual si t uat i on in which the const i t ut i on of homo sapiens and/
or t hat of fundament al physics itself is significantly different f r om what it
act ual l y is.
What emerges t hen is an unani mous silence on the details of the act ual
mechani sm of replacement. In part this silence is surely warrant ed since
it must be a mat t er of psychological investigation how observation terms
act ual l y get replaced or enriched, and whether piecemeal or by entire con-
ceptual schemes. On the ot her hand there is a corresponding semantical
di mensi on and we seem not t o have any but crude accounts of this - we
can choose between replacements tout court (and never mi nd about the
relations between successor languages) and some t ransi t i onal semantic
process (e.g. Sellars' analogical transference), possibly together wi t h the
t heory of observation terms as pragmat i cal l y condi t i oned responses t o
sensory nerve bombardment . Despite this unsat i sfact ory state of affairs,
the i mpor t ant positive definite t hi ng I wish t o emphasize is the Realist
criterion of success, namel y ul t i mat e theoretical domi nat i on of the ob-
servation language, in principle.
6.6. Realist Semantics ( V) : .4 Second Middle Way
6.6.1. Statement
In this section I am going t o briefly set down my own tentative selection
of doctrines f r om among those presented in the foregoing. It seems worth-
while t o do onl y because t hat selection differs a little f r om those I can-
vassed there and because it is i nt ended t o represent a comfort abl y sane
position (to be cont rast ed with the heroism of the extremes). I have labelled
454 C.A. HOOKER
it a second middle way.59 1 would not be overly disturbed to be persuaded
that, really, it was essentially a version of radical anti-Empiricism with a
mildly modified 'network model' of theories.
It should by now be clear that I agree with SeUars that, at least in prin-
ciple, theoretical terms can acquire an observational role and that theories,
ultimately, provide us with our truest characterization of the world. I
even agree with Feyerabend (and to the extent to which Sellars accepts
Feyerabend's cases, with Sellars) that many theoretical terms have already
acquired a reporting role in observation. I agree with Sellars that the
terms of the common sense framework are in an epistemologically fa-
voured position and I even agree that there is something semantically
distinctive about such observational terms. But here agreement stops.
6.6.2. A Distinction
First let me draw a distinction between a term's acquiring an observa-
tional role and its acquiring a reporting role in observation. The first of
these achievements I construe as follows: A particular concept achieves a
status such that the neurological processing structures that are the (cor-
relates of the) possession of that concept are habitually active in the per-
ceptual process, i.e. we normally process the incoming sensory information
using those structures, i.e. using that concept (among others). I have
spelled out this view of perception in detail elsewhere [58] but what this
amounts to is that we habitually (i.e. normally) conceptualize our per-
ceptions using this concept. The concepts of the commonsense frame-
work typically have this status. 6° By 'acquiring a reporting role in ob-
servation' however I understand only that we respond to perceptual
stimulation, in an appropriate set of circumstances, with reports (= de-
scriptions, claims, identifications etc.) that contain the concept in ques-
tion.
These two statuses are quite distinct, for on the one side there may be
ranges of circumstances where we habitually do not employ our per-
ceptually active concepts but only those that have acquired a reporting
role and on the other side concepts may acquire a reporting role in ob-
servation and yet play no natural part in perceptual processing. This
latter comes about through training and education. The most striking
examples are probably provided by working scientists. These people are
SYSTEMATI C REALI SM 455
educated to respond directly to laboratory situations with reports couched
in the technical language of theories. In fact this is exactly how the ' hard'
data, or at any rate, the prepared data, are reported, they are 'read'
straight off the instruments and this behavioural fact is what gives the
overwhelming impression that this is the only significant observational
level in science. There is no doubt that psyehologicaUy scientists do re-
spond directly in normal circumstances, i.e. there is no hidden psycho-
logical inference process involved which moves from the 'real' observation
reports, couched presumably in the 'real' observational terms, to reports
couched in theoretical terms. Nonetheless there is an important sense in
which even these people continue to perceive fundamentally in common-
sense terms even in those situations where they are not reporting in those
terms. 61 What has happened is that through training and education
these people are conditioned to respond directly to stimuli with reports
couched in terms other than those in which they fundamentally concep-
tualize those stimuli in perception. The situation is rather like (though to
be distinguished from) perceiving situations as theoretical states of
affairs.
This distinction is largely unappreciated in the philosophy of science. 6z
Once it is made we can see the elements of truth (but also that of error)
in both the empiricist and radical anti-empiricist extremes. Feyerabend is
right to think that scientists are not limited to commonsense vocabulary
for their observation reports but react directly to perceived situations
using the terms of theory because of a pragrnatie conditioning process.
He is wrong to believe that they leave the commonsense terms behind, or
that the observation experience changes conceptually, and for that reason
he is wrong to say that the observation language has changed tout court,
rather has it been enriched (and perhaps also been modified a little, of.
below).63 Moreover the grain of truth in the empiricist position is, rough-
ly, that we are stuck with the perceptually active concepts we have be-
cause these arise from the deeper structure of the world at large and our
particular human neurophysiology - and we retreat to this level whenever
we wish to make epistemologically less daring claims. But note that
epistemologieal caution does not equal indubitability, nor does it even
mean exclusiveness of observational role, much less exclusiveness of se-
mantical role, nor does it demand to operate at a phenomenalistic sense
data level - these are empiricist errors. 64
456 c.A. H O O K E R
6.6.3. The Nature and Role of Common Language
6.6.3.1. An untheoretical language. The distinction j ust discussed opens
the way to admit to science an observational vocabulary that is theory-
independent. Consider the perceptually active concepts; there is good
neurophysiological and historical-anthropological reason to suppose that
these have not changed significantly for millenia. (Aristotle was j ust as
capable of perceiving a shining orb in the blue sky as was Einstein, and
they would have agreed on that much despite their distinct theoretical
descriptions of the situation.) Moreover, a single scientist may switch
belief between totally disparate conceptual-ontological schemes, e.g. be-
tween field and particle theories, without any shift in his perceptions and
commonsense conceptual schema. These claims constitute strong grounds
for introducing such a level. Thus these concepts, and the simple general
truths that may be stated with them, are common between all the theories
of western science, they thus count as theory - neutral or independent in
this specific sense: 5 The terms of this language represent the perceptually
active concepts, they belong to the natural language (the common sense
level) and they have been with mankind for millenia, they remain as
theories come and go.
It is also true that this sub-language has become increasingly irrelevant
to science, because the concepts of scientific theories are increasingly re-
moved from the commonsense level. (Remember that only hard data de-
scriptions confront theory.)
6.6.3.2. Semantic ambiguity. Understanding the situation here is corn=
plicated by the fact that often the same sign has two (or more) distinct
uses, one (or more) at the theoretical level and one at the untheoretical
level (e.g. particle, space, exchange). Thus one particular token of the
sign-type may be heavily theory-laden, another untheoretical.
6.6.3.3. Theories and common sense language. Quite aside from the the-
ories that can be specifically formulated using commonsense concepts,
these concepts often have an additional theoretical load built into their
semantic content. 66 This is what partially justifies Feyerabend in referring
to common sense language as embodying false theories (in even his op-
ponent's sense of theory).
SYSTEMATI C REALI SM 457
6.6.3.4. Common sense language: is i t concept ual l y i nadequat e? You will
recall t hat the claim was t hat the common sense concept ual framework
clashed irrevocably with t hat of science. The implication was t hat it woul d
have t o be abandoned. I proposed denyi ng this latter conclusion. Let us
consider the case of a table, described as a solid wooden brown object and
perceived in corresponding terms. Accordi ng t o the usual account , the
common sense concept ual framework commi t s one t o t he existence of a
cont i nuous, coloured, impenetrable object with sharp boundaries, but in
fact none of these attributes is strictly t rue of the object before one ac-
cordi ng t o f undament al t heory, hence t hat object must be discarded in
t ot o as mere appearance. In fact, t hough, fundament al t heory and com-
mon sense agree t hat there is an object present at very accurately the per-
ceived l ocat i on of t he table, wi t h boundari es t hat are very nearly sharp,
which does in fact resist penet rat i on f r om ot her macro objects t o a very
hi gh degree and whose interstices are each very small compared wi t h
macro magnitudes. I f this were not the case it woul d be a very remarkabl e
world indeed in which t heor y coul d explain macro appearance and be-
haviour. Suppose t hen t hat we const rue t he common sense observation
report al ong these lines "Ther e is a table present which has no visible
interstices or breaks at its edges and which permits no perceivable pene-
t rat i on by ot her macr o obj ect s. . . " and const rue the furt her claim, when-
ever i t is present in the speaker' s meaning, t hat the table literally is inter-
stice-free, wi t h sharp boundari es and t ot al l y penetration-resistant as a
furt her (unwarrant ed) primitive expl anat ory at t empt t o be exorcized f r om
the commonsense scheme - t hen we obt ai n a eommonsense claim t hat
even f undament al t heory agrees is true. It is clear t hat the common sense
scheme is capable of drawi ng this distinction internally t o itself. I do not
claim t hat the ' purified' common sense scheme t hus obt ai ned must of
necessity be compat i bl e wi t h fundament al t heory in this fashion, 67 what
I do claim is t hat in our world we do have this compatibility. 6s But this
does not extend in quite the same fashi on t o the secondary qualities, for
I believe, eont ra Sellars, t hat objects are not e.g. literally col oured at all,
t hough t hey appear t o be. Judgment s of this ki nd must be exorcised f r om
the corpus; what need not be removed are the various t rue claims t hat
are instances of the general claim t hat objects possess properties in virtue
of which we sense objects in the qualitative fashi on we do. ~9 A very great
deal of the common sense scheme remains after these modifications, espe-
458 c.A. HOOKER
cially outside of the secondary qualities. (I have not discussed numerical
concepts, relational concepts, identity concepts and a host of other di-
mensions to the common sense scheme; in general all of these aspects
will be less in need of modification than those I have discussed.) All that
has been done is to render that scheme epistemologieally more cautious.
It is true that this has destroyed the rich theoretical and quasi-theoretical
conception of common sense objects tacit in the scheme, but it has left a
large portion of what we wish to say of them intact.
In sum, I claim that if we rid common sense language of its explicit
theories and the implicit theorizing built into semantic content we are
left with a theory-neutral scheme whose observational concepts are ex-
actly the untheoretical concepts of Section 6.6.1.; but these concepts do
not clash with theory, they are simply not very relevant as they stand to
specific theories from the point of view of testing them. Nonetheless they
are a fundamental part of observation and science. This neutral scheme
acts as a framework within which all members of a community may orient
themselves to one another, irrespective of their theoretical beliefs. Within
this common macro world the most important role for science which
these untheoretical concepts play is in the keying in of students of a theory
to the situations in which they are to apply the theoretically-loaded ob-
servation terms, i.e. in the process of acquisition of vocabulary with a
reporting role. Finally, the scheme has an historical importance as being
that out of which science grew (though its semantic role in that process is
obscure). Theory enriches this neutral observational vocabulary with its
reporting roles, it does not replace it, though in principle it could do so
(i.e. the appropriate changes in neuro-physiologieal perceptual processing
will be possible, in principle, if the theory is true - though the transition
is very probably not possible under normal human conditions). Theory
does, however, criticize and reject old common sense theories and beliefs
(where they are false) and likewise for common sense concepts having
false theoretical assumptions built into their semantics - here the clash
indeed occurs. Since these latter conditions are reasonably widespread in
common sense language there is a measure of truth in the usual view,
though the error made by attending to it alone dwarfs this truth in im-
portance.
6.6.3.5. The continuum of theory-ladenness. We find, then, as I earlier
SYSTEMATI C REALI SM 459
suggested, a roughl y cont i nuous progression from observational reports
couched in unt heoret i cal terms t hr ough the wide variety of ' hal f-way
house' terms t hat abound in common language to reports couched in the
technical terms of some specific theory. ( Common language is full of
terms, or part i cul ar word uses for a given sign, t hat are semi-theoretical,
largely t hr ough a slow osmotic transfer process from earlier theories. E.g.
metal, acid, vacuum, level. These terms come t o be integrated i nt o the
commonsense concept ual structure.)
6.6.3.6. Why untheoretical? There are t wo senses in which even our com-
mon sense perceptually active concepts involve a theory-like commi t ment .
To see what the first sense is, we begin again with the doctrine of per-
ception. That doctrine i nforms us t hat a logical prerequisite of perceptual
experience is the possession of a perceptually active conceptual scheme,
i.e. the possession of the appropri at e cortical processing structures. (Phys-
ically, the possession of these structures is a co-requisite of perception.)
But in a very general way the conceptual scheme which is perceptually
active embodies a metaphysics (it may in fact be capable of embodyi ng
several distinct metaphysical schemes). Thus the categories of perception
themselves will reflect a very general commi t ment t o a metaphysics for
the world. In our own case, for example, we habi t ual l y perceive the world
in terms of individuals located in a three-dimensional space and changi ng
in time, t hough the four-di mensi onal space-time world of relativity t heory
and the cont i nuous pl enum world of field t heory represent two logically
possible alternative metaphysical schemes which mi ght have been our
perceptually active scheme. In this sense t hen our habi t ual categorizing
in perception involves a very general commi t ment t o a view of the worl d
as made up in a certain way. Since I hol d t hat even general metaphysics
is ul t i mat el y open t o criticism in an exactly similar fashi on t o which the-
ories are criticizable (except for the time scale of criticism), this commi t -
ment can properl y be said to be a theory-like commi t ment .
There is a second, much more direct, sense in which the concepts of
the common sense scheme, afortiori those of the untheoretical scheme
(Section 6.6.3.5.), cont ai n a theoretical component , namely, it is simply
not possible t o teach those concepts or guarant ee the publicness (inter-
subjectivity) their social use demands unless t hey are underst ood (whether
tacitly or explicitly) t o funct i on in some laws (cf. e.g. Hesse [57], pp. 44f).
460 c.A. HOOKER
E.g. one only has to think of the network of geometrical laws in which
shape and size predicates are imbedded. The realization of this is anyway
immediate once one reflects upon the complex array of laws that must be
tacitly understood before even ostensive definition can be made to work.
(Pointing as a human act must be understood, the behaviour of sight in
relation to the body, the kind of thing being pointed out etc. etc.) 70 Most
of the laws involved are tacitly understood and seem only partially re-
flected in the semantics of the scheme (how deeply imbedded is the law
that a spherical body will roll on a flat surface? roll forever? - of. locally
euclidean but globally non-euclidean geometry), though the laws could
be stated within the commonsense frame if necessary. 71
It would appear that one cannot easily make a sharp distinction here
between tacitly held and semantically integral laws, nor easily between
theoretical extrapolation and untheoretical perception. To claim, e.g.,
that a spherical object will roll forever on a fiat surface seems clearly a
theoretical extrapolation from commonsense experience; to claim only
that an object rolls smoothly on the perceivable parts of a perceptually
flat surface to the degree it is perceptually spherical, so long as one is not
under illusion, is plausibly untheoretical. What this suggests is that strip-
ping away the theoretical extrapolation from the common sense scheme
may not lead to an obvious sharp demarcation of the untheoretical.
This is not a disaster because ultimately it will be recalled, the criterion
of the observationally untheoretical is that of active role in perceptual
processing. This latter can be determined, in principle, experimentally
(irrespective of the outcome on the theory-commitment issue). There never
was anything to guarantee that the stripping process for the commonsense
scheme outlined in Section 6.6.3.4. would yield exactly the observable
untheoretical concepts so defined or that these latter would forever remain
theory-neutral. 72
This discussion makes it clear that neither sense of theory like commit-
ment is logically different from straightforward theoretical commitment,
the difference comes in the different processes by which the commitments
came to be made (the one by a conscious, verbal process of theoretical
postulation, the other through unconscious neurophysiological adjust-
ment), and the different epistemological roles they play in the philosophy
of science (the untheoretical plays an ultimately dependent, but critical,
role in the inter-theoretic orientation of the human community). 7a Cer-
SYSTEMATI C REALI SM 4 6 1
tainly the untheoretical scheme, like the theoretical scheme, has been (and
probably still is being) elaborated over time in the light of experience,
though again by different processes. Like our theories, this elaboration
might go wrong (though it would have to do so in systematic ways to
allow survival). As I have said, one could imagine circumstances under
which fundamental theory informed us that many or all of our perceptual-
ly active concepts did not apply - we were under systematic illusion.
Nonetheless we have excellent reason to believe that nothing has gone
wrong in the process grounded on (i) general considerations concerning
what is known of the mechanisms of neurophysiological processing ad-
justment, (ii) historical evidence, (iii) the agreement of fundamental the-
ory with untheoretical commonsense. 74 1 regard it therefore as a reason-
able thesis that throughout the history of western science there has existed
an untheoretical conceptual scheme, that its observational component is
made up of exactly those concepts active in perceptual processing and
that it can be arrived at very closely by stripping eommonsense concepts
of their overt and tacit theoretical commitments. 75
6.6.4. The Semantical Content of Theories
I have no more informative account to offer of the sources of semantic
content for theoretical terms than that with which I concluded Section 6.4.4.
It does seem to me however that, bearing in mind the information-
processing model of conceptual activity, the imaginative roles of what
we call association of ideas, analogy, and very probably many transitions
for which we presently have no useful name, should not be under-esti-
mated in their contribution to a conceptual and, derivatively, a linguistic,
framework because of a precommitment to some 'origins' doctrines of
meaning where every concept has to be built up logically from some privi-
leged class of initial concepts. 76
6.6.5. Why Realist ? Recapitulation
The reader will recognize the foregoing account as a form of mild-radical
quasi-empiricist account of the observation language, with perhaps a more
liberal view of the possible sources of semantic content for theoretical
terms. 77 The compatibility of this view with Realism I have already ar-
gued. I simply remind the reader that to this position must be added the
epistemology and structural view of science of earlier sections.
462 c.A. HOOKER
7. A REALI ST ALTERNATI VE REJECTED" THE LEVELS HYPOTHESI S
In a recent, important book [40] J. C. Graves has defended a form of
Scientific Realism which I shall call the levels hypothesis. I shall endeav-
our to explain the view and then give my reasons for rejecting it as a
serious alternative open to the Realist. 7s
Despite Graves' rejection of Empiricism, and that of the epistemological
tradition in which it stands, the notion of a level for Graves is first an
epistemic one: Graves introduces cognitive levels (p. 17), only later do we
slip into ontological levels. The reason for this priority seems to be that
Graves thinks of levels as tied to theories and theories, Graves reminds
us, are obviously epistemic affairs (p. 19). Actually, the closest we come
to a definition is this: A level is the range of validity of a theory (p. 19).
This definition no doubt accounts for Graves' willingness to countenance
many different types of levels, e.g. that specified in the size hierarchy of
physics, the order of evolution of complexity (whether chemo-biological
or social) and so on (pp. 18-19). Intuitively, Graves seems to have in mind
that a theory (or at any rate a sufficiently general one - whatever t hat may
come to) typically either concerns a class of phenomena described in terms
more or less unique to itself (cf. molecular chemistry versus cybernetics)
or pretty clearly is concerned with the world only within some delimited
domain (cf. quant um molecular theory vs general relativity). Obviously
there is some truth to this claim.
Before pressing on from epistemology to ontology, let us pause to con-
sider the notion of a cognitive level. Even definitions using epistemic
components may not define epistemic entities. Such is the case with Graves'
definition of a level. For a Realist, the range of validity of a theory selects
an objective sub-division of the world; though a theory may be epistemic
qua constructed by man seeking knowledge, its truth conditions are non-
epistemic states of affairs (though of course any description we give of
them will be in the terms of some theory). Moreover, the definition is not
easily understood clearly, because no theory defines its own range of
validity. This is always a question of combining experience with mathemat-
ical meta-theorems (i.e. meta to the theory). But how experiments are
interpreted and combined with such mathematics is itself a matter for
theory, and not usually just the theory in question but others also, and
even an entire world view as well.
S Y S T E M A T I C R E A L I S M 463
Nonetheless, there is a certain plausibility about the not i on of cognitive
levels; we do seem e.g. t o know about tables and their putative at omi c
constituents in fairly different ways, or anyway different cont ext s - cer-
t ai nl y the descriptive languages are different. But t aken onl y t hus far the
idea is of minimal philosophical interest, it expresses no more t han a
rat her obvious reference poi nt for all phi l osophy of science. Not ori ousl y,
not hi ng follows for ont ol ogy from such epistemological remarks, for a
Realist. How t hen does Graves arrive at ontological levels?
By following the empiricist order of proceeding agai n: f r om episte-
mol ogy t o semantics. The different cognitive levels will in general have,
as I j ust said, somewhat different concepts associated wi t h them. Graves
takes this t o yield pri ma facie distinct ontologies for the various levels.
Graves concedes the possibility of reduct i on of one ontic level t o anot her
but offers a number of argument s for believing t hat many levels are and
must remain, ontologicaUy distinct. So then, the levels hypothesis turns
out not t o be merely an i nnocuous superficial description of science but
a serious claim concerning the structure of reality. Now we can reverse
the procedure and claim the ontological levels structure as groundi ng the
various episternic-descriptive levels. I shall not recount here all of the ad-
vantages Graves lists for this approach, none of t hem are guarant eed
uni que to the levels view, t hough certainly Graves is saved a l ot of work
on the relations among theories t hat someone who believes in onl y one
level is in for.
The alternatives which this hypot hesi s foists on one I find t ot al l y un-
acceptable in the light of experience t o date. Let us assume first P: No
two mat eri al objects can be in the same place at the same time, and second
Q: The spatio-temporal worl d structure consists of a single 4-dimensional
spat i o-t emporal manifold. Now consider any two ontic levels which
Graves believes distinct, e.g. the common sense level and the at omi c level.
Suppose we consider a table T and the corresponding collection of at oms
ti(i= 1, . . . , ,,~ 1 0 2 4 ) . Let the table occupy the vol ume V at time t and the
at oms respectively the volumes V~ at time t. Now i f the table T is ont o-
logically distinct from the structured collection of at oms T' , t hen either
P is false, or Q is false or R is false, where R is: Position measurement s
indicated t hat V, c V, most i. On the assumpt i on t hat R is an experi-
ment al t rut h, no one t o my knowledge doubt s t hat the structured collec-
t i on of at oms is located where the table is located (actually, all I need is
464 C.A. HOOKER
t he weaker ~ V1 : vi c V, which takes care of ' fuzzy edge' arguments)),
either P or Q must be false. But I can see no reason t o believe either t hat
we live in a worl d with multiple spaces of which we are all somehow
si mul t aneousl y conscious, or t hat t wo distinct objects can occupy the
same placetime, v9
Graves in fact offers t hree argument s f or t he claim t hat , irrespective
of t he consequences, levels must be consi dered ont ol ogi cal l y distinct, s0
None of these argument s withstands examination.81
I can t herefore see not hi ng to r ecommend Gr aves' position. I t woul d
appear t hat he is a victim of precisely t hat or der of priorities adopt ed by
Empiricism, which he criticized so sharply. ( How few confessed anti-
empiricists t rul y escape empiricist assumpt i ons ! - A caut i on t o t he reader
of this paper. ) The goal of Realism is not hi ng less t han t he const ruct i on of
t he t rue model of reality, a unified account of t he worl d, including its
reasoni ng inhabitants. Gr aves' pluralism of levels - cognitive, linguistic
and ont ol ogi cal - destroys t hat uni t y and t he mot i ve t o seek it.
8. ME T HODOL OGY
The t hree revol ut i ons (Section 1) are reflected also in met hodol ogy. Met h-
odol ogy I t ake t o be a br anch of pragmatics, roughl y it asks and answers
t he quest i on " Gi ven t he existing circumstances, which strategy best furt h-
ers t he devel opment of knowl edge?"
Empi ri ci sm const r ued ' t he existing circumstances' narrowl y t o include
j ust t he known facts and theories precedi ng t he time in quest i on (and it
const rued ' knowl edge' equally narrowl y t o mean a collection of pr opo-
sitions f or mul at ed in t he favori t e f or mal manner), s2 Feyer abend and
Hans on especially have at t acked this met hodol ogy vigorously. 8a Though
nei t her of these writers seems t o have explicitly f or mul at ed t he full-blown
alternative t o t he empiricist pr ogr amme adumbr at ed above bot h instinc-
tively at t acked the empiricist' s science and at t empt ed t o replace it with
a rat i onal account ori ent ed t owards scientists. In this cont ext the full
force of met hodol ogy as provi di ng a strategy f or increasing knowl edge is
evi dent - each scientist faces t he full cul t ure of his time and must act in
the light of all the likely repercussions of any given action. Feyerabend has
illustrated this poi nt brilliantly in his st udy of Galileo (and he was maki ng
poi nt s pri mari l y about epistemic strategy, not simply about Gal i l eo' s per-
SYSTEMATI C REALI SM
465
sonality, though we must begin back there for a full-blown explanation
of the actual course of science).
The methodological conclusion to be drawn from an epistemology of
ignorance, in the light of the revolutions in philosophy of science, is that
a vigorous pluralism is called for. When it comes to theoretical ideas ' Let
the hundred flowers bloom' . The reasoning is simple enough. Since
we begin in ignorance and with very limited sensory equipment, we have
only our native wit to help us to an imaginative conceptual grasp of the
world; experience shows that the most effective way to develop such
insights is in the constant critical play of ideas against one another.
Within the empiricist tradition one had such reverence for the domain
of 'facts' that they were endowed with large powers, e.g. the power to
point out the path of theoretical progress (if one countenanced such
at all seriously), but once that myth is abandoned in recognition of the
pitiful limitedness of our senses the way is open to concentrate on a
use where experience can be of moderately powerful service, as a
judge at the sensitive empirical meeting points of incompatible theo-
ries.
How vigorous the pluralism is depends upon other factors in one's
philosophy, s4 In my own case I am at liberty to exploit the approximate
sameness of the phenomenal level for all scientists, despite the possible
incommensurability of the hard data claims. But in addition I believe
that Feyerabend has seriously over-emphasized and over-simplified the
semantic disparities between general theories because of his neglect
of the complex roles of models and so-called auxiliary theories. Even
for our most general theories, sub-models of theoretical models may
coincide. More importantly, a powerful theory can model in its mod-
elling resources the models of another theory (thereby explaining why
they broke down and the degree of the failure). It will also generally
be necessary that incompatible theories share a range of auxiliary theories
in common.
Methodology is a branch of pragmatics for the Realist, its methods
dictated (in part) by semantics and epistemology; it is not necessary here
that I go into great detail. My own view is that every case of the clash of
theories needs detailed analysis as to the semantics of the two theories, the
range of models constructible and the relations between hard data reports
and phenomenal experience before any verdict can be given as to the
466 c.A. HOOKER
precise locus of the clash - so far as I can see, every (important) case
differs in philosophically important respects.
9. EP I S TEMOLOGY ( I l l ) : THE HI STORY OF SCI ENCE
With respect to the history of science we are faced with two extremes.
On the one side is the empiricist's account where we have an ever-accu-
mulating pile of facts with theories representing increasingly good orga-
nizations of them. Theories then supercede one another according to the
objective criterion of adequacy to organize the data, they are mutually
compatible where they overlap, indeed earlier theories are just for this
reason reducible to special cases of their successors. The history of science
presents a superbly rational front. At the other extreme alternative the-
ories determine their own semantic content and that of the observation
language and the domain of theory-neutral facts is entirely destroyed.
In this case we cut ourselves off from any rational account of the history
of science as a continuous process. For then a change of scientific theory
becomes a fundamental semantical shift of the entire vocabulary, it is
therefore impossible to meaningfully compare theories, and therefore im-
possible to give any rational account of progress from one theory to
another. Both extremes are unacceptable, s5
Ultimately, what account we offer of the history of science must depend
upon the account we offer of human beings. Both extreme views, in their
several ways, simply bring their a p r i o r i assumptions concerning ratio-
nality and epistemology (especially observation) to bear upon history
forcing it into the necessary mold. s6 Let me now elaborate briefly the
sources of continuity in the history of science which I have still allowed
myself, though we can expect the account to be elaborated and probably
modified as more information becomes available.
First of all there is the existence of an essentially untheoretical obser-
vational level which provides for a source of common experience to the
scientists of an entire culture and therefore for the scientific tradition
throughout an entire culture (e.g. that of western man during the last
2500 y). Since each theory must confront such elementary observational
experience, however tenuous this connection be, each theory can be com-
pared in point of its success in this regard. As we have seen, however,
such comparisons will not bear great weight in the decision among the-
SYSTEMATI C REALI SM 467
ories because of the necessity of the transposition of such experience into
a form suitable for direct confrontation with the theory in question, s7
Second, and at the opposite end of the scale, as it were, I have suggested
that there are broad metaphysical conceptions of the world, and corre-
lative to them broad conceptual descriptive schemes for the world, which
have been carried essentially unchanged through the entire history of
science. These very general conceptions of the world therefore also re-
present an element of continuity in the scientific process, ss Intimately
connected with such broad descriptive schemes lie certain broad mathe-
matical structures which also run continuously through the history of
science, s9
Within this limited continuity provided at the extreme ends of the
scientific structure, however, I believe we must admit that wholistic
changes in scientific theories have occurred through imaginative 'leaps' on
the part of individual scientists which can be followed only at the indivi-
dual level and probably cannot be given a rational reconstruction within
Science. Nevertheless, we should not under-estimate the force of the above
two sources of continuity, for I believe that such imaginative leaps as
there are have always arisen out of a detailed controversy concerning
certain specific characteristics or defects of an existing theory 9o and the
particular solutions which have been adopted have always been adopted
with an eye to conforming to the fundamental metaphysical-conceptual
schemes of the culture and (as far as possible) to the fundamental mathe-
matical structures of the preceding theory. To what extent the theories
may he compared in detail on this view of the history of science, i.e. to
what extent the meanings of theoretical terms are theoretically determined,
and to what extent there is any common sharing of ontologies between
competing theories, we shall leave as open questions here. My general
conclusion is that the history of science is one of continuous creation
under the careful constraints of a rationally reconstructable continuity.
Finally I raise the question of whether these culture-wide aspects of
science cannot themselves in some way be opened to critical examination.
I suggest the view that ultimately both of these aspects are open to the
same kind of critical reappraisal as are any other elements in the structure
of science - it is just that the reappraisal time is so much longer than for
other elements and the reappraisal process so much more complex and
difficult to realize. 91
468 c. A. HOOKER
In respect of the perceptual level it is difficult to imagine t hat our per-
ceptual organization would have even this much flexibility, precisely be-
cause it is so dearl y determined in large measure by our neurological
make-up, and we have reason to believe t hat this was acquired only after
millennia of evolutionary experience. Nevertheless, there is reason to be-
lieve that this neurological make-up has some degree of flexibility built
into it (of. [54]) and one cannot help but at least consider it possible t hat
over the course of a much longer period of time (perhaps millennia) the
conditions under which man operated, cultural and perhaps also neuro-
logical, could so change as to alter his fundament al perceptual organiza-
tion, i.e. to alter the concepts active in perception and/ or the manner in
which they function together. Indeed there is talk of an ' evol ut i onary
epistemology' [21], i f so it will be a naturalistic epistemology, one whose
course is critically underst ood within science.
POSTSCRIPT
The reader will recognize t hat it is not possible to do more here t han
provi de a sketch of many of the aspects of a systematic Realist position -
in particular, this is true of an account of scientific discovery and scientific
change, reduction and the unity of the sciences, the nature of values and
their roles in the sciences and the relations between the pursuit of science
and cultural activities generally. For the same reason the notes often con-
tain more substantive material t han their conventional role would justify.
Even so, the space this essay occupies is substantial and I wish to express
my gratitude for the generosity of Prof. Jaakko Hi nt i kka and Synthese.
NOTES
x Thus there is no insistence here on identification of theory with a linguistic formula-
tion in some used language (whether common language or that of the scientist), mathe-
matical-logical formulation of a theory is acceptable, the theory being identified equally
with the set of axioms set down or with the class of intended models of those axioms
(i.e. physically irrelevant models are eliminated).
Equally, I do not assume that the Realist is concerned only with theoretical entities
(i.e. individual substances) but admit theoretical properties also, nor assume that these
latter are predicated only of theoretical entities, nor observable properties only of non-
theoretical entities - cf. Achinstein [1], Putnam [I03], Sellars [117] etc. But this is only
to forewarn the reader of issues to be discussed in what follows.
2 One of the problems arising here is the so-called 'incommensurability' of world views:
SYSTEMATI C REALI SM 469
by definition, a truly compl et e worl d view contains a cohering versi on of everyt hi ng
(ethics, ont ol ogy, epistemology, linguistics, science etc.), thus one can never ' get out-
si de' it, there is no neutral gr ound; but in this case we all seem doomed t o forever be
unabl e t o even discuss issues seriously wi t h our opponent s (and vice versa), since such
discussion, ff it is t o be a genuine meeting of minds, must be on neutral ground. For -
tunately, t he following condi t i on seems adequat e t o guarantee rat i onal debate wi t hout
presupposing neutral ground: t hat each worl d view be ri ch enough t o faithfully re-
present within itself opposing world views. Our ability t o talk about ot her st ruct ures
(logical, mathematical) using a part i cul ar structure, ot her languages using a part i cul ar
language, is wide indeed. However, this possibility does not by itself guarantee t hat
decision among world views will be rational, or determine what the basis of that de-
vision is. I have hinted here and in [73] t hat a st rong ethical component is involved.
a Feyerabend and Kuhn have been greatly mi sunderst ood (and perhaps have even
mi sunderst ood themselves) in this because al most everyone operates with the pri mary
not i on of Science. Fr om this perspective one accuses t hem of confusing psychology
and sociology with epistemology and met hodol ogy, or of introducing t ot al irrationality
i nt o Science. But once one views their wor k from t he new perspective t hen much of
what is puzzling falls i nt o place. E.g. Kuhn' s account of paradi gm-domi nat ed science
is t o be seen as articulating the t ot al epistemic cont ext f r om within which decisions in
some historical peri od ar e made and Feyerabend' s account of Galileo-as-propagandist
is t o be underst ood as a rat i onal account of ' gambl i ng with t r ut h' in an ideologically-
epistemically l oaded situation etc. Cf. my remarks in [73].
4 The t erm ' new epi st emol ogy' derives from a paper of G. C. Nerl i ch' s which I saw
duri ng t he summer of 1972 and which proved very helpful in clarifying some of my
ideas. Al ready a year earlier t han that I had been independently thinking about t he
pr obl em as e.g. my bri ef remarks in [64] indicate. About t hat t i me I re-discovered t he
i mport ance of Qui ne' s ' Epi st emol ogy Nat ural i zed' (in [108]). I am also indebted t o
John Ni chol as f or some further ideas t hat arose out of discussion of a paper of his in
Oct ober 1972.
6 A characteristically clear earlier statement of this position is found in Qui ne' s ' Epi st e-
mol ogy Nat ural i sed' (in [108]). Accordi ng t o Quine, the abandonment of t he Empi ri -
cist semantical and epistemological pr ogr amme for science (respectively dogmas 3 and
4) leaves no task remai ni ng f or t he t radi t i onal concept i on of epistemology and it must
either be abandoned (as e.g. t he l at er Wittgenstein woul d have) or reduced t o a branch
of psychology, specifically t o t he quest i on of how in fact creatures with our make-up
arri ve at our beliefs. Similarly, Qui ne sees no rol e for t he t radi t i onal approach t o se-
mantics, but in this case advocates its abandonment in favour only of a t heory of
reference and truth. (In similar fashi on we mi ght construe Quine and Davi dson - pace
Ne r l i e h- as suggesting t hat t he abandonment of the traditional approach t o t he nat ure
of t rut h shifts philosophic interest from a pointless issue t o t wo real probl ems: how to
speak coherent l y of t rut h - a probl em solved by Tarski - and how t o account for the
knowledge of infinitely many truths when a finite language is employed.) I n my vi ew
though, Qui ne overstates t he case, for he makes it sound as i f all critical reflection
upon science is t o be abandoned in favour of j ust science. I surely believe, al ong with
armchai r philosophers, t hat one has t o distinguish t he critical appraisal of science f r om
science, otherwise t he l at t er becomes blind. Quine on t he ot her hand, seems bent upon
excluding normat i ve considerations entirely. Qui ne is right t hat the criteria of episte-
mically justified action merge much mor e conspicuously i nt o t he general cri t eri on of
rat i onal action but, I think, t oo hasty in droppi ng the normat i ve di mensi on of the
470 c . A. HOOKER
latter. And it is not true that normat i ve theories cannot be i nformed by experience.
What is really needed is a dynamical t heory of the interaction of normat i ve and de-
scriptive theories as experience changes - but t o supply that here woul d lead us t oo far
afield. Quine is not qui t e right, t he key question is not simply ' How do men in fact
theorize in response t o st i mul at i on?' but something closer t o ' Gi ven creatures of our
construction, in our circumstances, including the answer t o Qui ne' s question, which
parts of science are epistemically justified? How should the rat i onal man respond
theoretically?' While on t he subject of this i mport ant essay, let me gather together one
or two ot her remarks concerning the views expressed in it.
Quine, as always, is unrepentantly empiricist about language, about meani ng espe-
cially. The only place we have anything remotely satisfactorily identifiable as meani ng
ar e the so-called observat i on sentences and their meani ng is stimulus meaning. But by
his own lights, Qulne should surely wait upon the deliverances of psychology in this
area ! For all we know, a rich not i on of meani ng will be fort hcomi ng t hat departs radi-
cally f r om stimulus meaning, it mi ght be t hat the human mi nd has an inner predilection
and facility for the abstract. Even Qui ne' s own not i on of stimulus meani ng must be
supplemented by internal stimuli i f it is t o make any sense of the distinctions between
an observation report uttered in veridical, illusory and hallucinatory circumstances (a
poi nt I made while reviewing a thesis, and then discovered it made independently by
Shick [ 115] ) - who can tell apriori how ri ch these inner stimuli are? My own remarks on
semantics which follow show my Qui nean uneasiness with the not i on of meaning,
but I am unwilling t o write it off (or down) in his fashion until mor e is underst ood
about t he relations between language, t hought and act i on - cf. Not e 68.
Quine is also quick to poi nt out how under the new epistemology t he not i on of an
absolute observat i on sentence loses its sense t o be replaced by some not i on concerning
nearness t o stimulation in the causal chain. I t hi nk this must be right. But his qui ck
step from there t o a Feyerabendi an pragmat i c t heory of observation where what count s
as an observation sentence depends upon no more t han the gross behavi our of t he
linguistic communi t y is j ust t oo quick. It ignores what psychology may tell us (to doubl e
Qui ne' s i rony - p. 88), and this may be i mport ant in relation t o the view of theories
taken. For example, i f psychological research suggests a permanently fixed fund of
observational concepts with no condi t i oni ng possible then Feyerabend will be simply
wrong. If, as I believe - cf. below - we must distinguish sharply sensory cont ent f r om
observat i onal reporting, wi t h condi t i oni ng essentially affecting onl y the latter, t hen
Feyerabend has t oo simplistic a view and the quasi-empiricist Ramsey account is in
some difficulty. Because it suggests none of this detail, Qul ne' s vague proposal is un-
helpful and uninformative.
s Though not destroy, of course, the insight t hat a man might adopt t he right beliefs
for t he wrong reasons. But t he standard account makes the bridging of the dualism
in a fashion which does reasonable justice t o the history of science al most impossible.
7 I am indebted to discussions with J. J. Leach on these issues - the reader interested
in this approach t o i nduct i on is invited to consult his [83], as well as Luckenbach [85].
s Cf., e.g., the discussion by Maxwell [90], [91].
9 Roughl y speaking, these correspond respectively e.g. t o Hospers' concept empiricism
and judgment empiricism - cf. [75], Chapt er 2.
10 The reader is referred to the standard literature for a statement (and appraisal) of
these doctrines - see, e.g., Achinstein & Barker [4], Aune [8], Ayer [9], Carnap [23],
[24], Feyerabend [27], [28], Hempel [47], [48], Nag¢l [92], Pap [95], Qui ne [106], [107].
Cf. Lewis [ 8 4 ] .
SYSTEMATIC REALISM 471
11 This remark is explained a little later on. Historically, it is probably the case that
explicit Conventionalism is a late arrival, being considerably pre-dated by Instru-
mentalism (perhaps better called Fictionalism) which played a dominent role in the
discussions of the significance of astronomy from Ptolemy onwards. Cf. e.g. the remarks
by John Winnie in an as yet unpublished paper ' Realism Without Regrets'. Winnie
also discusses the historically operative auxiliary motives for support for Instrumen-
talism; roughly, they come down to this: Instrumentalism can be invoked to nullify
any theoretical view that one has independent grounds (empiricist, religious, ideolo-
gical, etc.) for wanting nullified.
Originally, I had intended to deal only with the various particular kinds of conven-
tion that might appear in a physical theory, in a later section. (I still do this, see my
paper [68], Sections I, ID. Prof. J. O. Wisdom convinced me that to do only this would
be an error. Furthermore, the expression of the conventionalist position (though not
my basic understanding of it) is indebted to a recent discussion of Wisdom' s-see [136].
Though I do not follow his solution, his discussion contains many illuminating remarks.
Further discussion of Conventionalism can be found in the references of Note 8 an
in Wisdom [137].
12 But first let us be absolutely clear that Realism (my Realism) concedes the initial
premise of this general attack. Prescinding for the moment from the question of how
the actual structure of science is best represented (further on which see below), let us
consider the traditional (Empiricist, Conventionalist) structural theory:
Level Type ":i [ ~"
(I) (x) (... T...)
(2) (x) (... T, O...)
(3) (x) (... O...)
(4) . . . O, a. . .
The structural relationships hold among them, in general, as the arrows suggest. Level
(1) contains only initially universally quantified expressions containing only theoretical
terms. (That is, I assume the initial quantifier to be universal, although it may be fol-
lowed by one or more existential quantifiers; actually, I have some sympathy for the
view that we could possibly find a fundamental law essentially expressible only with
a first existential quantifier but here I need only the conclusion that the kinds of laws
stated do occur in science, and in fact they predominate.) This first level is the level of
' pure theories'. (Not all theories have such a level and it is not clear what ' pure' involves
here, but I let that pass for the moment.) Level (2) has similarly quantified expressions
that mix observational and theoretical terms. These supply the so-called ' correspon-
dence rules'. Level (3), while similarly quantified, contains only observational t erms-
the observational generalizations. Level (4) contains the unquantified sentences reporting
' facts' . In this standard account level (3) is deducible from levels ( 1) + (2) and some
members of level (4) from other members together with level (3). Each level is imagined
deductively axiomatized and the whole likewise. Fact: only members of level (4) (and
perhaps also a very small and uninteresting sub-class of level (3)) are sense-confirmable
(i.e. can have their truth values ascertained on the basis of sensory experience). (The
members of level (3) which I have in mind might be assertions of the form ' All the
marbles in the palm of John' s hand are red' , where ' marbles' is understood as speci-
fying an object at least one centimeter in diameter.) Conclusion: no interesting level (3)
sentences, and afortiori no level (1) or level (2) sentences at all, are seuse-confirmable.
Logical truth: no inference from any collection of level (4) sentences to any collection
472 c . A. HOOKER
of level (3) sentences is ever deductively valid. (Even if the level (4) sentences in fact
exhaust a class of beastics, that particular fact is not reported among them.) Fact: no
instance of a theoretical predicate is ever observable (qua theoretical predicate - but
cf. Note 1). Conclusion: not even a single instance of level (1) and level (2) sentences is
sense-confirmable. Afortiori, a fortiori level (1) and (2) sentences are not sense-con-
firmable. There is no escaping this conclusion. Fact: the increased expressive richness
of levels (1) and (2) permits an infinite variety of theories to deductively imply the same
subset of sense-accessible level (4) sentences and hence be ' empirically equivalent' .
(Notice that this is a loaded, anti-Realist conception of empirical equivalence. The
Realist will say only that they are observationally equivalent.)
Now the Realist is apt to reject the strict observational/theoretical dichotomy on
which this structure is based and from which its epistemology flows. Even so, he will
be hard-pressed to reject, as generally adequate, the hierarchy of logical form posed
(skipping the observational status of the predicates occurring in the theories) and hence
cannot reject the problem of the invalidity of the inference from instance to generaliza-
tion. (In [56] Mary Hesse writes as if the whole structural description could he rejected,
I suppose on the grounds that the ' netwoIk model' has no use for such hierarchies,
cf. [57], but this is surely to exaggerate. True, epistemologically loaded hierarchies
such as the observational/theoretical, or concrete/abstract will not follow without
further epistemological assumptions from a structural description, but it is implausible
to reject the formal deductive hierarchy since analyses of actual theories suggests that,
by and large, it is appropriate (see the appendix of my [71] for an instance where pure
deduction doesn' t suffice going ' downward' ). The empiricist analysts (Caruap, Hempel,
etc.) certainly were familiar with formal analyses of scientific theories.)
However, in addition I at least accept that there is an important element of truth in
the empiricist observational/theoretical dichotomy (though I do not by any means
think they have stated this correctly), and so I, and more generally anyone who is pre-
pared to concede that theories even often deal with unobservable entities, must concede
the general epistemological conclusion drawn above from the traditional structure (see
below for a detailed discussion of the alternative Realist position here). This agreed
feature of the epistemology of science is the basis for the attacks of the opponents of
Realism.
18 I shall present here only the briefest recapitulation of the attacks on Empiricism,
my aim being to show how they presuppose alternative world-positions.
Psychologically, Locke' s account of concept formation is plausible only ff one ne-
giects the hereditary structure of the central nervous system and its powers, allowing
the mind only logical operations with which to transform the only ideas (concepts) it is
allowed, namely those given directly in sensory experience. This is the power of the
view - it is also its great empirical weakness. The view of concepts inherent in the doc-
trine assumes them to be conscious, linguistic units, whereas a more adequate view
recognizes the notion of information-processes in the central nervous system as the
fundamental units - in this latter case one obtains a more effective account of concept
formation and a very great enrichment of the category of concepts. (Some of the rami-
fications are lightly spelled out in my [74], cf. Note 64.) The only dispute among psy-
chologists today is how rich the hereditary structure is (less rich it seems, than Kant
thought it was) and how the fantastically rich variety of imaginative transpositions
occur. (See, for example, [41], [43], [77], [97], [102], [139] in bibliography; for a more
detailed discussion of perception see my [58].) And Realism takes science, hence psy-
chology, seriously.
SYSTEMATI C REALI SM
473
As a bol d statement of the appropriate analysis of the psychology of percep-
tion, the dogma 3-dogma 4 combi nat i on is found similarly want i ng (there is none
but superficial evidence for the existence of simple uni t s, givens, i n perceptual ex-
perience - c f . my [58], Chapter 4 and the references cited there), and as a charac-
terization of the history of science dogma 4 is agai n i nadequat e (plenty of obser-
vat i onal claims have been revised i n the light of experience, see especially the
studies by Feyerabend [27], [28], [29], [32]), yet a sufficiently ' hard-nosed' empiricist
mi ght insist on ret ai ni ng dogma 4 as an in-retrospect and in-principle analysis of per-
ceptual claims and of the progress of science (using some subtle theses concerning
descriptions t o count er the objections - cf. Section 6.4 below).
Philosophically, doctrines 2, 3, 4 and 5 have come under attack. There are powerful
argument s to show that, cont ra P-Empiricism and Empiricism dogma 3, theoretical
terms as they occur i n actual theories of science cannot be defined i n terms of purely
observational vocabulary. (Cf. the works by Aune [7], Nagel [92], and Pap [95] and
Sneed [123]. One of the ki nds of argument s that does not work for (or respectively
against) Empiricism is the argument from the formal separability of the observational
from the theoretical part of science, for it presupposes the rejection of Realism (re-
spectively, Empiricism) - cf. my [59], [60].) These argument s are ' ni ce' because they
largely hinge on features of ext ant theories t hat any plausible view woul d want to i n-
corporate. The honest poi nt is, polemically, t hat to offer an interpretation of a theory
is not simply to show how its heretofore uni nt erpret ed terms are connected to ' si mpl e'
observation, but to say what is really going on i n the world as opposed to, but ex-
pl anat ory of, what appears to be, is observed to be, going on. One does not want
to know merely t hat i n circumstances C and i nst rument dial reads R but what is
really going on between the i nst rument and the world such t hat it comes to read
R. This clearly involves theoretical ont ol ogy but it also involves other things, e.g.
a theory of perception, which Empiricism ignores. Nonetheless our hard-nosed
empiricist mi ght cling to his reconstruction of science as the onl y ultimately justifi-
abl e one.
(Recently Maxwell has claimed t hat Russell' s doctrine of the knowledge by acquai n-
tanee/knowledge by description dichotomy, bolstered by the application of his theory
of definite descriptions to a Ramsey approach to theories (for Russell' s doctrines see
[112], [113], cf. Maxwell, [89], [90], [91], and for Ramsey' s approach see [109], cf. Sec-
t i on 6 below for a detailed examination) reconciles concept Empiricism with Realism,
for in this case we use onl y observation terms and logic to refer to theoretical entities.
More precisely, Maxwell has dogma 3 al one i n mi nd, he rejects dogmas 2 and 4 and
would reject dogmas 1 and 5 1 believe, and onl y a W-Empiricist version since not every
t erm is reducible wi t hout remai nder to observation terms, specifically not the described
theoretical terms. Tentatively I believe this claim to be correct; tentatively, because the
Realism that results is so pallid I am not sure it can st and - of. Section 5.4. If so, what
this demonstrates is that i f you cut one of the empiricist dogmas off severely enough
from a systematic view and weaken it sufficiently it becomes compatible with almost
anything.)
The Verifiability Criterion of Meani ng which is the general semantical support for
the foregoing const ruct i on has come under severe attack and has now generally been
abandoned. (This is essentially part of the at t ack on dogma 3, see [1 ], [4], [7], [9], [15],
[23], [24], [47], [48] [49], [50],[51], [57], [92], [95], [100], [103], [104], [105], [I06], [107],
[116], [122], [126], [129], [130].) Indeed the Security Dogma has the characteristic t hat
it onl y needs to be clearly stated for its implausibility to be evident. The Epistemological
474 c . A. HOOKER
Dogma is somewhat weaker (and subtler for this reason) but its support also comes
from the same basic empiricist doctrines - indeed, it is t he ki nd of doctrine f or which
none ot her t han this kind of support is imaginable.
Recently t he legitimacy of the observational/theoretical distinction has been attacked
(sec Achinstein [1], Put nam [103]), but not, I think, effectively against a ' hard-nnsed'
opponent. (The argument here is parallel to t hat immediately below. Cf. also Suppe
[125].) Qui ne and Put nam, among others, have attacked dogma 2, with uncertain sue,-
cess. (Quine [106], [107], Put nam [105]. Qui ne' s argument seems to hinge on the claim
t hat i f any of a collection of terms can only be inter-defined (i.e. defined in terms of
each other) then t he whole collection wants for sense. (So much the worse for language l)
Why should not an empiricist introduce his preferred language, already compl et e with
a sharp analytic/synthetic distinction (as Carnap wants to) and then lay down explicit
syntactic definitions of these semantical categories as a f ai r accompl i ? (The extensions
of these semantical terms can be claimed to approxi mat e - and merely sharpen - our
preanalytic intuitions.) Besides, it woul d seem that Qui ne' s preferred remai nder to
semantics (truth, reference etc.) is in no better position on this score (and on some
others, cf. my [61].) Moreover, it is arguable that Quine also adopt s an implausibly
strict (almost Positivist) cri t eri on of what can count as part of t he definition of mean-
ings etc., namel y overt behavi our alone. However, ff one adds verbal and internal
stimuli t hen a defensible position may result - for a recent defence, cf. Wilder [135].)
Dogma 5 (Phenomenalism) has been consistently attacked recently (s¢0 S¢llars [116],
Armst rong [6], Feyerabend [29]) but it seems to me that a sufficiently hard-nosed
Phenomenalist coul d legitimately reject the arguments since they all hinge on particular
characteristics of our present conceptual structure more highly t han t he doctrine of
Phenomenal i sm itself (cf. my [58]).
Above and beyond these detailed criticisms, there is a quest i on of meta-consistency:
t he basic tenets of Empiricism (and of Conventionalism) constitute a general t heory
about the nat ure of man and his rel at i on to t he world, they seem neither reducible to
' sensation and reflection' (Empiricism), nor to conventions, i f they are to make the
sense their proponent s require of them. Empiricism will need to show its own tenets
ne~,essary truths, or offer some transcendental deduct i on of them, and convent i onal i sm
make sense of its own implicit general account of language, before these doctrines
achieve the coherence requi red of a world view.
14 The instrumentalist can hardly claim the fact t hat theories are used i mport ant l y t o
make inferences as distinctive support, the Realist agrees t o t hat - it is t he claim that
they are onl y instruments to this end that needs support.
ts In this connection, the new work being done on a ' systems epi st emol ogy' or an
' ecol ogy of knowledge' t hough often misleadingly formul at ed is the beginnings of a
mor e explicit integration around the historical spiral of epistemology with social science
l ooked for in Section 1 - see e.g. Bate.son [10], Piaget [98], Wojciechowski [138], and
of. my [73].
i s Realism is incompatible wi t h t he semantics of phenomenalistic reduct i on certainly,
but in this case we have something more direct - the ont ol ogy of psychology taken
Realistically is not compatible with the ont ol ogy of Phenomenalism, nor do the actual
psychological details support the Representative Realist theory.
1~ One perfectly legitimate analysis of physical theories suggests t hat quant um t heory
demands an objective revision of the laws of logic - cf. my [61]. I happen not to accept
the conclusion in this case f or ot her reasons, but I mi ght have, had these ot her reasons
not applied.
SYSTEMATI C REALI SM 475
For a stimulating departure from foundational epistemology in science see Feyera-
bend [31], cf. my [62], [63], [70].
is But an epistemologist with my persuasion would be inclined to take a similar view of
historical theses, religious doctrines, etc.
l~ Certainly for Ds anyway. If a semantical presupposition of the use of our kind of
language turns out to be some very general facts concerning the world and our neuro-
logical structure, then the condition sets in D1 may in fact partially overlap, or even
overlap quite a lot, though remaining semantically distinct kinds of conditions.
These distinctions ought to be refined by the further distinctions I drew in [64], p. 167.
so There is, e.g. ontological simplicity, conceptual simplicity, simplicity of axiomati-
zation, simplicity of derivation of key theorems and simplicity of representation. This
last includes Schaffner's notion of fitness (cf. [ 114]). For an in troduction to the complex-
ities of simpllcity see Rudner [111], Bunge [17], Foster and Martin [35]. On the role
of regulative ideals in science see also KOrner [78], Nash [94], Toulmin [131], [132].
sl In case anyone is so naive as to believe that metaphysics is not essentially employed
in science they should reflect upon the very general principles governing specific theo-
retical realizations of such ontological schemes as atoms-and-the-void. Cf. my [65] and
[72] and Bunge' s proto-theories, Bunge [15]. CL also Agassi [5], Bur r [18], Koyr6
[79], [80].
ss See Section V. As to speculative metaphysics' being thought to be unrelated to the
scientific enterprise I can only say that, just as in science, it is a pragmatic decision
(perhaps culturally mediated) to exclude it, and like all pragmatic decisions one runs
some risk of being embarrassed. One can only defend the decision on as reasonable
grounds as possible (the lack of evidence for the relevance of the concepts to science
etc. - none of the arguments are very strong, nor can they always even be very clearly
expressed at present). Thus there remains only a small sieve where once the Positivists
had thought to erect an impenetrable barrier.
ss See Popper [100], especially appendix viii, but cf. also Not e 33.
34 Since I have just finished stressing the lack of a fixed first philosophy in my Realism
this is the place to add that it is obviously the case that Realism as I conceive it is
logically independent of Materialism - I conceive of Materialism (and its alternatives)
as scientific theories to be decided on the same merits as any such theories; one only
obtains some form of Materialism if one adds to Scientific Realism some thesis con-
cerning the exhaustiveness of say, physics. Similarly, Realism as I defend it is a far cry
from the good old anthropomorphically epistemological days in philosophy when man
felt sure of himself and Realists made bold claims to know the necessary structure of
reality - I have no truck with this brand of apriorist metaphysics.
Illustration: Given what I have said here, and about induction earlier, I could hardly
opt for a ' necessary connection' doctrine of the causal relation in anything like the sense
Hume attacked. Roughly speaking, I opt for a Hurnean-style account of causal rela-
tions (and of laws generally, since ' cause' is not a very important category in mathe-
matical science); causal laws are generalizations derivable from theories rather than
from spatic~temporal relations among initial conditions, they have as their basis energy
exchanges and it will always be some such basis as this that allows us to distinguish
them, it is their being derivable flora a theory that leads to our saying that they support
counter-factual conditionals. I doubt that we are ever likely to be in the epistemological
position to confirm theories that introduce non-extensional causal relations against
purely extensional versions, but if we are so I should happily accept such a strengthened
doctrine; in the meantime the Humean-type account looks the most acceptable.
476 c . A. HOOKER
On this approach, cf. also Braithwaite [14], Nagel [93], Hempel [51], Hanson [44].
~ Thus despite his careful exami nat i on of the position - and the related Ramsey
transcriptioulsm, cf. Section 6 below - and despite his referring to my [59], where I
stated this argument 5 years before, Cor nman still manages to miss the poi nt ; he con-
centrates on the question of specific defects i n the approach because he takes a narrow
concept i on of the positions involved instead of l ooki ng to their global dimensions.
se E.g. one might want to defend onl y the weaker view i f one thought, as I do, t hat it
is an empirical question whether nat ur e is a cont i nuum with respect to any physical
variable (e.g. space, time, spin etc.). Cf. [42].
Suppe [125] presents a version of this argument.
s7 A not untypical empiricist move would be to l op t hem off the theory proper and
locate them i n an ' accompanyi ng model ' which acts as a heuristic device but does not
ent er the structure of science essentially. Cf. Nagel [92] and Sellars' nice discussion i n
[117]. Cf. also Suppe [125], [128]. On the roles of models i n general see also Campbell
[22], Harr~ [46], Hesse [54], especially see her [53], Henki n e t al. [52], McMul l i n [86],
Suppes [129], [130].
ss I n r e s p e c t o f c o h e r e n c e i n s c i e n c e , o b v i o u s l y a t h e o r y is i n c o m p l e t e l y internally
g l o b a l j u s t t o t h e e x t e n t t h a t it is e x t e r n a l l y g l o b a l . T h e ideal, totally i n t e r n a l l y g l o b a l
t h e o r y w o u l d e n c o m p a s s all o f science.
F o r t h o s e i n t e r e s t e d i n t h e f u r t h e r e l a b o r a t i o n o f t h e s t r u c t u r e o f s c i e n c e o u t l i n e d
i n t h i s e s s a y a n d e x t e n d e d i n [71], R u n G i e r e h a s p r o v i d e d a n i c e d i s c u s s i o n o f t h e
structure of the statistical inferences involved i n an epistemological setting closely sim-
ilar to mine. See [39].
29 Bunge [16], and Suppe [125], [126], have independently discussed a not dissimilar
structure for science and various others have discussed fragments of it, e.g. van
Fraassen [36], [37], Hempel [47], Hesse [57], Sellars [117], Shapere [120], [121], Suppes
[129], [130]. (I have had the opport uni t y of seeing these works by Bunge and Suppe
onl y i n the interval between the writing of the penultimate and ul t i mat e drafts of this
paper.) CT. also my remarks i n [71].
8 o I n t h i s connect i on Maxwell has emphasized t hat an i mport ant part of the charac-
terization of the general conditions of an experiment will be the obt ai ni ng of t heoret -
i cal l y characterized circumstances (e.g. thermal or phase equilibrium, negligible angul ar
moment um pert urbat i on etc. etc.), These conditions will i n fact i n general be supported
by the auxiliary theories (and confirmed through their dat a support) and by whatever
supports the cet eri apari bus clause - though Maxwell does not go i nt o this detail, What
Maxwell emphasizes, rightly, is the damage this realization does to any simple verifi-
cationist or faisificationist approach to science. See [91].
I n the same paper Maxwell also makes the poi nt t hat often a scientist will not know
the exact hypothesis that will relate a theory ' properl y' (i.e. i n the fashion one intuitively
expects, given the metaphysics etc., see below) to a range of experimental situations
but will guess that there is such a relation and go on to test the theory on that basis,
hopi ng t hat those very tests will reveal the correct relation. I t is an interesting and i m-
port ant poi nt because, though seldom made, science abounds with such moves. I woul d
onl y add t hat what guides i nt ui t i on here are the general metaphysics and prot o theories
operative plus the models suggested by the theory and its auxiliaries. As Maxwell
remarks, this complicates the confirmation rel at i on still further.
ax E.g. there will be calibration of instruments and det ermi nat i on of errors; also the
t ransformat i on of readings i nt o model-properties. This last, like the former, introduces
addi t i onal theoretical assumptions, e.g. the construction of a smoot h density funct i on
SYSTEMATI C REALI SM 477
from a finite mass sample, of a velocity and location from time-of-flight measurements
and so on. All of these apply equally to human sense organs as to machines.
s2 Here, e.g., we must correct the system behaviour for the influence of factors ignored
in the theoretical model and also make the transition to any further idealization built
into the theoretical model - point masses etc.
as Thus Hempel' s deducihflity conditions for explanation I take to be necessary, but
not sufficient, conditions of a satisfying explanation. It further follows, I believe,
that, although all sound explanations of this type will also provide sound predictions,
mere prediction is a much weaker requirement, so that not all predictions that pay
off need rest on a basis sufficient for providing a satisfying explanation. On the deductive
model of explanation and the explanation/prediction asymmetry see, e.g., Hempel [51],
Suchting [124]. On the role of models in science see the references of Not e 27.
s4 Sometimes, for example, scientists argue after a form ' If T and E then with prob-
ability p (near 1) S, so assume S; now if S. . . ' and sometimes they argue ' i f S then T,
assuming T we shall also assume S: now if S. . . ' when predicting how things will go,
as well as when assessing the degree of support a theory possesses (of. Not e 30). Of
course many far more complex argument forms than this appear as well in actual prac-
tice. In many, many appfications of a scientific theory one cannot say for certain, given
the theory under examination, precisely what the experimental conditions axe, though
one can usually know very probably what they are. Unlike the empiricist structure,
therefore, which segregates deductive and inductive arguments, the present model per-
mits both types of argument to occur in all theory-data transitions.
a5 By ' model theoretic' I mean assertions couched in the descriptive terms of the model
and not a formal metamathematical language.
s6 These themes are expounded in my [65]. For some earlier intimations, see also [64],
Section 13. Further Realist exposition is found in [66] and [67].
s 7 I might add that, even narrowly construed within the stage 2 development of induc-
tion (Section 1) and hence as a specialized part of stage 3, the foregoing analysis re-
quires a radically re-vamped theory of confirmation. Actual falsification, equally with
verification, can hardly be expected to be realized (cf. also [91], and Note 40). But
confirmation itself must be seen as something considerably different from the theory
of the direct data sentence - theoretical sentence moves suggested by the simple tradi-
tional structure to science; confirmation may eventually turn out to be a purely logical-
style relation, but the characteristics one wants now to give it leave this issue in doubt
to say the least. First, theories are decided among, and confirmed to differing degrees
by the same evidence on quite other criteria than purely deductive ones, criteria which
include ' fit' with preferred metaphysics and proto theories, simplicity, fecundity etc.
Second, thereis in general no simple relation between a theory and either ' its' hard data
or ' its' phenomenal data, transitions to each require auxiliary theories, ceteris paribus
clause, model corrections etc., and may often not even be deductive - the support or
lack of it that reports formulated at either level furnish a theory is indirect and me-
diated by a number of considerations concerning all of these other factors (e.g. the
likelihood that, given the preferred metaphysics, the ceteris par&us clause is O.K., etc.)
(Moreover, the very language in which observation is reported is dominated by the
relevant theories and so the reporting may be only as reliable as the theory to he re-
ported on - cf. also [57], p. 61.) Third, the degree of support a given experimental out-
come furnishes a theory is at least partially determined by that theory itself, because
it supplies part or all of the theory of the instruments used l Thus the theory specifies
how accurate an instrumental reading is likely to be, whether it directly reflects the
478 c . A. HOOKER
theoretical paramet er value etc. and so determines t he degree of support the instrument
readi ng yields that very same t heory (cf. my [71] for details). For these reasons t he
simple empiricist not i on of confi rmat i on as a quasi-logical rel at i on between dat a sen-
tences and theoretical sentences must be abandoned t o be replaced by somet hi ng that
is closer to the actual structure of scientific reasoning - what that will be I do not know.
Equally, it seems to me t hat by and large the goals of verification and falsification are
ideals seldom, i f ever, at t ai ned; they may function as regulative ideals, they can hardl y
act as demarcating criteria for the scientific (cf. also [91] - anyway this circumstance will
hardl y bot her us; cf. Section 10 for remarks on methodology).
ss E.g. the belief that solid things are wi t hout interstices - it is not that we see t hat
objects are interstice-free, it is j ust that we do not see any interstices. Not i ce that this
is a different position from Sellars who so closely links concepts and beliefs t hat the
phenomenal level must be scrapped in t ot e in the final analysis. Cf. my discussion in
Section 6 and in my [66].
s9 E.g. it does not include the concept of a poi nt mass, this concept can be made
rigorously consistent t hrough the t heory of distributions, it does include ' line of force'
because these are intended to be individuals and yet according to t heory there woul d
on occasion have t o be real, i rrat i onal numbers of them.
40 One of the weaknesses in the present work of Suppe is that he does not distinguish
these vari ous cases.
41 See Maxwell [89], [90]. Ramsey' s original idea was advanced in [109] and l at er
developed especially by Bohnert [12], [13]. I have comment ed upon it and cited some
more literature in [59] and [60]. Cf. also Cor nman [25] and the mor e recent literature
he cites.
42 The only thing we ar e supposed t o know about t he theoretical predicates is t hat
their extensions are interrelated, and related to t he extensions of observat i on terms,
in certain ways. This suggests a further, more extreme, move: drop the explicit occur-
rence of bound variables ranging over properties, and pass instead to bound variables
ranging over j ust the extensions of these unknown terms (so that predication in effect
becomes membership). This extreme version remains parasitic upon the concept of a
theoretical predicate, however, until such times (ff ever) as a direct translational tech-
ni que can be formul at ed for passing f r om the untranscribed t heory t o the purely set-
theoretic transcribed version in a way which avoids using some ki nd of comprehensi on
axi om to define the sets and insure their regularity that woul d not commi t one t o t he
theoretical predicates. I f this severer formul at i on could be made t o work, however,
I believe it would better capt ure MaxweU' s quasi-empiricist structuralism, cf. below in
the text. I am indebted t o Ni ne Cocchiarella for helpful criticism at this point.
43 Cf. my discussion in [59] and that by Bolmert [12], [13]. The difficulty with the l at t er
view is that we do not yet have any precise account of inductive argument forms and
so one cannot produce a clear-cut answer as to whether devices exist to cope with them.
Cf. the discussion by Comman [25].
44 The reasons for the unlikelihood are these: f or the non-theoretical object satisfying
t he Ramsey claim t o have only the properties which the observable level of t he relevant
theory admits is for that t heory to be reducible t o observational statements al one ~ la
P-Empi ri ci sm and for t he properties t o be observable but not so reducible is highly
implausible (how did we miss them?), yet for the properties to be theoretical is certainly
t o re-admi t a significant sense of theory, even ff it is restricted t o an implausible macro-
phenomenal i st ontology.
The actual structure of Ramsey transcribed sentences is considerably mor e compl ex
SYSTEMATI C REALI SM 479
t han most accounts of it (including Maxwell' s) suppose. For some of these complexities,
see e.g. Sneed [123].
4~ Actually, in the form in which Maxwell presents the argument it is made to com-
preahend not onl y theoretical terms in the usual sense but also the ordi nary first or der
properties of the macr o world, thus arguing for some version of Lockean representative
real i sm; but we can distinguish the two applications, resist Maxwell' s extension of the
argument and defend a version of direct realism i f we wish. This latter coarse is the one
I woul d mysel f argue for, indeed I see it as essential to a viable Realism, cf. above and
[58], and so in my account here I shall construe this argument as applying onl y to
theoretical terms. Not e that this restriction is compat i bl e with the Realist rejection of
the observational/theoretical di chot omy in the strict empiricist sense and its replace-
ment by a theoretically defined distinction within psychology.
4e Though I agree with Sellars [117] that Nagel [92] does not succeed in reconciling
Instrumentalism and Realism, the force of part of Sellars' attack on Nagel may be
vitiated by the foregoi ng remarks since it is not clear that Nagel really intended it t o
be logically impossible f or theoretical terms to ' descend' to t he observational level.
47 It should surely be clear t hat to claim t hat X is actually realized and t o claim that
X is only physically possible (i.e. t o claim that X' s occurrence is merely compat i bl e
with the true, fundament al laws of physics whatever they be), is t o claim distinct things,
afortiori, to claim merely that X' s occurrence is compatible with the laws of logic is to
make a third distinct claim.
48 These moves are as follows:
First, they can make use of an observat i on/ report i ng distinct/on t o poi nt out that
scientists ar e condi t i oned t o respond t o experimental situations in theoretical language
j ust as they respond to mor e ordi nary situations in more ordi nary language, j ust this
may account f or the felt familiarity; indeed, I see no objection t o the inclusion of such
responses in a quasi-empiricist account of meani ng (' el ect ron' is t he sort of t erm I use
to respond t o situations of . . . sort), in this case the t erm mi ght even be said t o have a
connot at i on of sorts, t hough not one directly concerned with the subject mat t er of the
theory.
Second, I believe they coul d adopt Sellars' view t hat theoretical terms share hi gher
or der predicates in common with terms havi ng a connot at i on (cf. [117] and Section
6.4.3. below), they might even insist on this t oo as a component in genuine connot at i ons
for theoretical terms.
Third, they mi ght poi nt out that the metaphysics of a t heory already specifies the
ki nd of t hi ng or propert y t o which many theoretical terms ar e t o refer, i f the met a-
physics is grant ed a connot at i onal cont ent it coul d partially confer it on these terms.
One could easily add t o the Ramsey transcribed t heory a collection of assertions giving
this addi t i onal cont ent (e.g. t hat the t erm electron stands for a class of objects such
that t he members in its extension share the attributes mass, mot i on, locality in common
with billiard balls and such that it is used by scientists when respondi ng descriptively
to cl oud chamber photographs), and I do believe that this woul d come a l ot closer t o
grasping the cont ent of a t heory t han does the unadorned Ramsey approach, but it
involves i nt roduci ng a meta-linguistic component to theories and t o my knowledge the
appr oach has not yet been extensively explored.
Of course, t o t he extent t hat all of these sources are t aken as sources of genuine
connot at i on, t o that extent the strict Ramsey analysis must be abandoned (since none
of the claims that would fol l ow analytically from the connot at i ons thus acqui red fol-
l ow f r om t he bare Ramsey treatment). Mor e seriously, this account is worthless until
480 C. A. HOOKER
it is specified how ordinary language and the terms of the metaphysics (if distinct)
acquire their connotations. If the answer is: through observation, then the latter he-
comes impossible to understand (how could a metaphysics possibly be rich enough to
specify a general world ontology for a variety of theories if its terms are confined to
observational content? - unless it is done purely denotationaUy, in which case this
treatment for theories themselves will surely be sutfacient) and the former leads us back
into the dogmas of Empiricism (not to mention the lack of an account of how attribute
sharing can enxich the language beyond the original observational resources, cf. 6.4,3.
below). But if the source of connotation is mysterious we have failed to advance at all.
On the other hand, the ' pure Ramseyite' might simply wish to insist that all of these
sources contribute not to genuine connotations for theoretical terms, but to psycho-
logical states in scientists in which they feel the normal familiarity with the terms.
(Here observation as the source of connotation may be safely retained.) This move
requires an analysis of the distinction between meanings and the psychology of language
use that supports this possibility. I find it less attractive because I am persuaded that
the foregoing sources belong to a proper account of the meanings of theories, I also
find the difficulties of the alternative unappealing; but until this approach is explored
and found decisively wanting it seems a consistent way out.
49 In the former case, the history of such ' terms' as molecule, gene etc. which, we would
normally say, once referred to unobservable entities that are now - thanks to modern
instrumental technology - observable, poses more problems for this position. The only
consistent answer is to deny that, in some strict sense, these entities are observable even
to.day. We shall see later that this might be done, without falling into the trap of es-
pousing a phenomenalist analysis of sensory experience, but the Realist here is forced
to tip toe along a razor edge between disasters. (The modified Ramsey approach seems
distinctly preferable on this score anyway.)
50 Indeed, the notion of ostensive definition is quite compatible with some formula for
meanings of words in terms of the set of rules governing their usages, since the osten-
sivist can claim either that a sensorially handicapped person could not nndelstand the
meaning of all of the rules associated with an ostensively defined term (i.e. could not
understand a rule of the form ' Apply this predicate when you arc in these circumstances',
circumstances demonstratively indicated) and/or would not be able to apply all the
rules suceessfully. It is important to understand all of the logically possible combina-
tions of doctrines here in order to appreciate the reasonable room to manoeuvre which
the Realist has.
In fact, a form of the two arguments used earlier to support this alternative can be
employed here to argue for the necessity of ostensive definition. To quote Mary Hesse:
...there must be a stock of predicates in any descriptive language for which
it is impossible to specify necessary and sufficient conditions of correct
application. For i f any such specification could be given for a particular
predicate, it would introduce further predicates requiring to be learned in
empirical situations for which there was no specification... We must, there-
fore, conclude that the primary process of recognition ... is necessarily un-
verbalizable. The emphasis here of course is on primary, because it may
be perfectly possible to give empirical descriptions of the conditions, both
psychological and physical . . . . but such descriptions will themselves depend
on further undescribable primary recognitions. (Hesse, pp. 39-40, author' s
itafics.)
SYSTEMATI C REALI SM 481
(Those who would deny the hietm~hy of analysis appealed to here and envision instead
a language in which, circularly, every predicate would have its conditions of application
described in the language, only succeed in 'smearing" the ostensive support throughout
the language, not in dissipating it - cf. Note 13.)
~1 For Feye~bend' s doctrines see, cg., [27], [28], [32], - for a more complete biblit>.
graphy together with a critical review see my [63], and for specific critical comments
see, e.g., Achlnstein [2], Butts [19], Putnam [104], Seliars [117], and my [70].
s~ The only other possibility, that theories have their semantic content conferred from
some external source, does not seem open to Feyerabend, for there are only alternative,
and hence semantically incompatible, theories that could act as this source, or else it
is completely mysterious - the level of neurophysiological irritation is common to all
theories and hence of no use here. Feyerahend certainly does not discuss the possibility
of an external source and the structure of his philosophy suggests that he would dis-
count it.
5a Roughly, Beth defines a k-ary predicate K(xx .... xD to be explicitly defined relatively
to a set of formulas A and the predicates they contain besides K if there is a formula
U containing k free variables and other predicate letters from A such that
(X1)... (Xi') (K(xl ..... xJ:)*-+ (J(Xl, ..., xt))
is entailed by A, and implicitly defined if
(xl)...(x~) (Kt xl . . . . . x k ) ~ K" (xl . . . . . x~))
is entailed by A U B, where B is obtained from A by replacing each occurrence of K
by some other k-ary predicate K" which does not appear in A. See Beth [11], p. 290.
I have van Fraassen to thank for reminding me of the importance of this theorem.
54 To take a hopelessly trivial case, ff in the postulate (x) (,,ix ~ Bx) the concept of an
,4 is that of membership in a set of things which include the set of B' s as a subset then
this postulate is already an effective tautology and it becomes so twice over ff the con-
cept of a B is that of a set of objects included in the set of A's.
55 Not that the former would automatically rule out the position - given Feyerabend' s
assumptions it seems to me that a case could probably be made out for the view that
a bunch of tautologies could act as an effective guide to life. After all, they only have
to be so related to conditioned behavioural respones that people survive, continue ex-
perimenting etc. and such relations do not seem ruled out by the tautological character
of the beliefs. Cf. on this score the functioning of a myth - see e.g. Feyerabend' s com-
ments in [31]. Those who automatically assume that tautologies cannot he a guide to
life either tacitly assume an empiricist account on which tautologies are both restricted
to logic and uninformative and/or assume that only by conscious assent to the cogni-
tive content of a belief, as making a difference to prediction, do beliefs affect conduct,
but this ignores the subtleties of the conditioning role of language. Nonetheless no one,
to my knowledge, has tried to grasp and develop this interesting thorn in detail and so
I too shall leave it resting on its rose vine with only this brief defence.
On implicit definition see also the excellent discussions by Frege [38], Hempel [50],
Nagel [92]. Cf. my discussion in [70] and Lewis [84].
5s But Sellars concedes that for Feyerahend the only interesting feature of theories is
their direct connection to sensory responses, Feyerabend ignores this independently
described subject matter in all cases, so he is able to construe natural language as a
theory, in his sense - [117], pp. 172--4.
6~ See [117] pp. 178-184 and the references cited there. At a crude level of approxima-
tion the complex relations holding between theory and the common sense framework
which are mediated via the models may be viewed as the traditional correspondence
482 c . A. HOOKER
rul es; cf. t he theory-phenomenal correspondence rules in my structural model in Sec-
t i on V above.
b8 Certainly there ar e some features which strike one as immediately right about this
sort of account. Pretty clearly our ancestors developed language initially as stimuli to
action and these sounds, laboriously built into a systematic pattern, can he supposed
t o have been t horoughl y behaviourally oriented. But the actual developmental his-
t ory is a bit hazy, and insight is not assisted by the fact that we know of only ' compl et ed'
languages (i.e. languages we can map ont o our full-blown commonsense framework).
Specifically, Sellars does still rely quite heavily on analogical enrichment of t he obser-
vat i on-t o-t heory part t o get us up-hill from a purely observat i onal base - onl y the invalid-
ity of construing an analogical rel at i on as a logical one saves Sellars from the empiricist
semantical dogma (i.e. dogma 2, t hat all descriptive terms are analyzable in terms of
pure observational terms). But t hen what exactly happens t o semantic cont ent during
such analogical shifts, how it is transformed, how the mi nd has t he capacity to carry
out this operation, and how we achieve a ' l eap' t o richer concepts remai n somet hi ng
of a mystery. And there are mor e specific difficulties as well. (See here the exchange
between Marras and Sellars [87], [119], [88], which it seems to me Marras has won -
at least at this stage of the debate.)
Moreover, there is in Sellars still the ol d empiricist insistence that observat i on brings
somet hi ng special to concepts. For t he additional semantical richness by which theore-
tical terms at t ai n t o full stature is j ust their use in observat i on once again. In addition,
I cannot see why observat i on (or its internal equivalents) in the common sense con-
ceptual scheme are held t o confer a special, irreplaceable semantic cont ent as Sellars
seems t o believe. (See the detail argument in my [58].) Not e t hat my attitude does not
entail t he rejection of the not i on of ostensive definition.
~9 The ideas in Sections 6.6.2.-6.6.3.6. are expounded in [66].
60 Not all of t hem do because the common sense framework itself is shot t hrough with
theories, cf. below and [66] e.g.
61 Thus ask a scientist ' what is t hat ?' poi nt i ng t o an object, and he, assuming a tech-
nical question may reply ' I t is a Wilson cloud chamher ' ; but i f you then say ' ar e you
sure?' in a fashion which suggests not merely technical ambiguity but t he possibility
of a mor e radical error, he will immediately ret urn towards his perceptually active con-
eepts, he mi ght say e.g. ' Wel l it looks like a cloud chamber, there' s t he cylinder for t he
plunger and here t he count er ar r ay. . . ' - t he harder he is pressed t he further towards
puri t y in t he use of these concepts he will be driven.
68 Bot h Popper and Feyerabend recognize elements of t he psychology of percept i on
which backs this distinction, though wi t hout putting t hat t heory t o work in the phi-
losophy of science very pointedly, cf. [30] and [101]. Hesse [57] recently discussed t he
cont i nuum of epistemological caut i on within science wi t hout introducing the psy-
chology of percept i on; otherwise, much of her comment ar y makes excellent sense.
Schaffner [114], pp. 323-4, comes close t o this distinction (he draws a distinction nearly
t hat of my Co, C concepts distinction in [66], t hough he makes it theory-conflict re-
lative) but smothers it with semantical restrictions (e.g. t hat there are no experimentally
gratuitous conceptual distinctions) and a narrow view of it determined by its alleged
history.
This distinction is properl y encompassed within an analysis of ' seeing that' (rather
t han ' seeing X' ) which, as I say, I have offered in detail elsewhere [58]. Suppe [125],
e.g. is correct in sensing t he i mport ance of such an analysis for t he general philosophy
of science, but fails, as do most of his colleagues, t o take t he psychology of percep-
SYSTEMATI C REALI SM 483
t i on seriously enough and so misses t he actively-observing/reporting distinction.
e3 The fact is then t hat t he psychology of perception is not t aken seriously enough
when discussing the not i on of observat i on in science. I f to this lacuna we add the view
t hat the connot at i ons of terms are pragmatically determined, i.e. determined by their
uses t o which their users have boon socially condi t i oned, t hen we indeed collapse t he
distinction between report i ng and t he actively observational. This is j ust how Feyera-
bend et al . seem t o have overl ooked t he distinction. But not e t hat even with t he prag-
mat i c det ermi nat i on of connot at i on granted, t he distinction remains once an adequat e
psychology of perception is introduced.
e4 The distinction is also not much discussed in phi l osophy proper, i.e. within the
philosophy of perception in epistemology, which is, I intuit, the reason for the blindness
in philosophy of science. Philosophical doctrines of percept i on have tended t o operat e
within t he confines of the empiricist-phenomenalist analysis of percept i on where per-
ceptual reports were simply exact descriptions of some conceptualized content, a sense
dat um; there was an assumed identity of cont ent and report . But in fact perception is
neuro-physiologically and linguistically much mor e compl ex (and interesting) t han
this. What we report of perception, as indeed the cont ent of our perceptions, ar e func-
tions of the cortical processing undertaken. Nor are the linguistic report age and per-
ceptual processings totally tied together (indeed, it is al most certainly not the case that
all conscious perception is linguistically conceptualized - cf. my [74], [58]).
What I have in mi nd here is the i nt erpret at i on of experiments on people whose right
and left brai n lobe connections have been severed. See [93 ]. Since onl y one l obe possesses
linguistic ability and yet bot h l o b ~ respond as t hough fully conscious, what ar e we to
say? It s ~ ms plausible t o say t hat so l ong as neurophysiological processing is rich
enough t o provi de all t he usual perceptual i nformat i on (as evidenced by behavi oural
skills etc.) then, in effect, the usual perceptually active concepts ar e active here as well;
but there is no correl at i ve linguistic ability.
I n any case, this view seems reasonable j ust on the ground that the i nformat i on-
processing view of the brai n seems likely t o prove the most adequat e one and the
possession of concepts is most plausibly tied t o the possession of i nformat i on-pro-
cessing cortical structures; mor eover we know that in a normal brai n t he localized
speech centre is onl y a small por t i on of one lobe. What this suggests is that perceptual
skills and, i f we accept t he argument of the preceding paragraph, conceptual skills are
much broader t han linguistic skills. Indeed, i f we coul d onl y rid ourselves of the supersti-
t i on that t hi nki ng and seeing were kinds of internal linguistic acts we should be well
on t he r oad towards saying somet hi ng sensible about non-cognitive communi cat i on,
artistic creativity, supra-linguistic mat hemat i cal activity (and symbolic activity in gen-
er al . . . ) and even ' bei ng stuck for words' . For furt her el aborat i on see my [74].
e~ It woul d take a l ong argument to extend this cl ai m to ot her cultures as well, but
this I woul d want t o do by emphasizing three points: (i) t o possess a concept is funda-
mentally to possess an information-processing structure in t he cortex, hence concepts
are expected t o be determined by experience and to be richer t han language - cf. Not e
64, (ii) what language skills a culture develops depends upon its social interests etc.,
i.e. it is a partly pragmat i c affair, (iii) at best one mi ght expect some conceptual-per-
ceptual schemes to be fragments of others, not incompatible with others, for there is
no evidence of massive illusion in any culture. Usi ng these moves I would at t empt t o
render the ant hropol ogi cal work on ot her cultures - e.g. that by Whor l [134] on the
Ho p i - compat i bl e with the cl ai m in the text.
n6 E.g. t he Oxford English Di ct i onary defines solid as (i) impenetrable etc. and (ii)
484 c . A. HOOKER
completely filled up etc. Pretty clearly semantic component (ii) began life as a primitive,
perhaps unconscious, but anyway false, attempt to explain (i).
07 I can conceive of a world in which science reached the conclusion that unaided man
was under systematic illusion in every respect (that science was possible would depend
upon the pattern of the illusion), though I do not believe it probable.
6s This holds as true of the esoterica of quantum theory and general relativity as of
general atomic theory. Unless geometry were locally euclidean, unless macro scale col-
lectivities of micro entities do not fluctuate in state over perceptible time intervals, etc.
we could not plausibly explain our macro-world. There is here no fault with the guarded
macro observation reports I have delineated (though some theoretical additions to these
have to go).
e9 We have to remember that the perceiving of secondary qualities as exemplified by
objects is deeply ingrained in our evolutionary structure for obvious good reasons, it
represents a quasi-theoretical commitment (Section 6.6.3.4.) rather than a theoretical
one. For a detailed treatment see [58]. Though even within the common sense frame-
work there are reasons to believe that objects do not literally exemplify the secondary
qualities (the strongest reverse case is that of sight with the other senses in varying
degrees less bothersome in this respect), I believe we must concede that on any occasion
on which we literally attribute secondary qualities to objects that we do so falsely and
our corresponding perceivings are illusory.
Sellars places heavy emphasis upon the secondary quality aspects of the common
sense conceptual scheme and suggests that to modify them in the fashion I suggest is to
abandon the scheme. I do not believe this is necessary. (The concept of, and conditions
of, illusion can themselves be discussed in common sense terms, nor are qualities neces-
sary, cf. [58], [69].)
70 Thus the earlier arguments for ostensive definition (Section 6.4.) must not be under-
stood as claiming that certain terms, those ostensively defined, must be understood
before any others can be understood - this is clearly impossible in most cases. (Some
elementary terms, e.g. ' Dad' , might be understood prior to the emergence of some
fragment of a conceptual scheme.) Rather a conceptual scheme emerge* slowly in a
child' s head with many terms taking on clearer' definition' - cf. Not e 63. In this sense
the ' smeared' model of ostensive support is very probably the more accurate.
71 These particular dimensions to common language give the final sense in which
Feyerabend is right to hold that common sense language embodies a t heo~, though in
this case it is less than certain that it embodies a false theory. For an elaboration of this
doctrine see my [70].
v~ If it could be shown experimentally that the elaboration of the untheoretical core
was at the behest of contemporary theories, rather than general features of the macro
structure of the world, then my position would become indistinguishable from a (rather
' wild' ) version of radical anti-Empiricism.
If we discovered that the elaboration of this core was influenced by theories held at
historically earlier times then we should no longer have a coherent account of how
science was possible, though ff we discovered (more plausibly) that the common sense
linguistic scheme was progressively infected by byegone theories then we should have
one more reason to distinguish concepts in general from linguistic concepts and to
distrust the linguistic semantical dimension.
78 It is necessary to keep these points clearly in mind, otherwise one will be tempted to
run backward along the line of evolutionary advance searching for a purer and purer
realm of untheoretical concepts. This search is chimerical, for as we regress the un-
SYSTEMATI C REALI SM 485
theoretical conceptual scheme becomes cruder and more fragmentary, far enough back
we run out of concepts altogether.
74 Indeed, to quote Mary Hesse:
...the comparatively stable area within which it is proposed to define an
observation language itself is partly known to us because its stability is
explained by the theories we now accept. It is certainly not sufficiently
defined by investigating what observation statements have in fact remained
stable daring long perioda of time, for this stability might be due to accident,
prejudice, or false beliefs. Thus any attempted definition itself would rely
upon current theories and, hence, not be a definition of an observation
language which is theory-independent. Indeed, it might justly be concluded
that we shall know what the most adequate observation language is only
when ~f possible we have true and complete theories, including theories of
physiology and physics which tell us what it is that is most 'directly ob-
served'. ([57], p. 49.)
The difference between Hesse and I is that she thinks that this truth renders the un-
theoretical/theoretical distinction insignificant for science whereas I do not, for the
reason given. She also thinks that it leads irresistably to the ' network model' , whereas
I do not - cf. below.
75 The idea that there might be no hard and fast observational vocabulary for science
is not a new one. When discussing one particular realization of this idea - the so-called
' network' model of theories - Mary Hesse offers two pages of references in fine print
to others in which this idea, or fragments or analogues of it, may be found, beginning
with Duhem and Campbell, though it is clear that earlier work in the nineteenth century
already contained the issue ([20], [22], [26], [133], [136], for Hesse' s bibliography see
[57], footnote 4). And Schaffner derives his views on a model similar to my own from
Popper. Nonetheless many of these accounts (though by no means all) suffer because
they discuss the issue in isolation from a psychology of perception and concept forma-
tion. The network theory is a case in point. The position which I have tried to develop
attempts to self-consciously take into account these domains of psychological theory -
in particular, I believe that any realistic view of the function of observational vocabu-
lary in science must draw a distinction between the reporting and perceiving roles and
link that distinction to the distinction between perceptually active concepts and those
not so, further that the relative stability of the common sense observational vocabulary
must be explained in terms of this former category of concepts. The network theory
removes certain semantical problems by declaring them chimerical but seems to ignore
psychological structures behind theorizing (it ignores the distinctions I have drawn).
The reason for this is that the passage is made too hurriedly from the denial of the
analytic/synthetic distinction, via the network model, to the denial of the observational/
theoretical dichotomy (observation sentences are simply those ' on the periphery' ). I
will not dispute that the denial of the analytic/synthetic distinction leads to the network
model but one only moves from thence to the denial of any epistemologically signifi-
cant observational/theoretical dichotomy if one also tacitly assumes that the psychol-
ogy of perception is either irrelevant to this issue or supports the pragmatic theory of
observation terms - I hold this tacit premise false. (This argument shows that the re-
jection of the analytic]synthetic distinction is not sufficient for the rejection of the
observational/theoretical dichotomy; pretty clearly it is not necessary either. It follows
that its rejection is not sufficient for Realism, and I believe not necessary either, at
486 c . A. HOOKER
least for the quasi-empiricist forms.) However, Hesse and Schaffnvr do in fact come
quite close t o my position, Hesse when she concedes t hat as a mat t er of fact some predi-
cates are more stably observable t han are others (of. [57], pp. 50-1), and Schaffner
when he allows t hat one may ' anal yze back' observation statements t o increasingly
' deeper ' levels, until a level of agreement in meani ng between conflicting theories is
f o u n d - [114], pp. 321-2. (Hesse also makes nicely clear t hat the net work model, though
it encourages t al k of convent i on, is quite distinct f r om convent i onal i sm; [57], p. 43.)
7e The cort ex is capable of a tremendously compl ex range of activities, there is every
reason not to hol d it down to some simple combi ni ng operations on preformed con-
cepts (as e.g. empiricists woul d have us do). If t o possess a concept is to possess a
cortical processing structure of some sort then we can expect there to be a myri ad of
ways in which these can be built up, ways which quite transcend our attempts at se-
mant i c description. How, after all, di d we come by our ' nat ur al ' language? Will some-
thing like Sellars' Ryl can account do? What was the exact relationship between the
psycho-cortical and social processes at work and t he emerging semantic cont ent ? The
semantic description seems t o be relevant, so to speak, only t o the relatively definite
recognizable linguistic eod-products which result from these fantastically compl ex pro-
cesses. It seems impossible t hat every psycho-cortical or social act occurring in t he
langnage-acqulsition process has a semantical correlate (especially when we recall t he
non-linguistic powers of t he cort ex - cf. Not e 64); we reco~miTe semantically relevant
features of language only after a sufficiently rich and definite linguistic framework has
developed. It is when I t hi nk thus t hat I can believe t hat Quine was fi ght t o suggest that
there was somet hi ng inherently groundless or chimerical about meanings (i.e. conno-
tations), t hough only in this sense, cf. [107]. (Not e that t he foregoing is a small part of
t he response t o t he arguments f or quasi-Empiricism presented in Section 5.2.)
77 It mi ght be helpful at this stage t o briefly present t he view of the nat ure and rol e of
observat i on concepts in science as I see it in order to distinguish it from bot h Em-
piricism and mor e radical Realisms.
Af t er t he fashion of the Empiricisms I admi t to t he occurrence of relatively theory-
i ndependent observat i onal descriptive concepts. I have tried t o say why I believe this
t o be t he most reasonable alternative.
But t he empiricist produces his observational real m as a mat t er of a priori analysis;
which concepts are observat i onal is fixed a priori and hence permanently, consequently
the observational/theoretical di chot omy is also fixed a priori and so permanently. By
contrast my doctrine grows out of experience together with a psychological doct ri ne of
perception - which concepts are observational, even in the theory-neutral case, is deter-
mi ned by the facts of our world, including our neurological make-up, and is in principle
open to change as those parameters are vari ed; for this very reason the observat i onal /
theoretical di chot omy is itself a theoretical distinction. Precisely for this reason it is
not , as Feyerabend woul d have [30], a completely pragmatically determined distinction.
Nor is t he distinction theoretically dependent in the sense that its application in t he
domai n of some t heory is dependent on that t heory as Hesse suggests ([57], p. 49) unless
perhaps for the t heory of neurophysiological evol ut i on itself and like theory. Bot h
of these furt her claims ignore t he observat i onal / report i ng distinctions (these further
claims are true of terms that have acquired a reporting role). (Hesse offers us the ex-
ampl e of the concept ' is simultaneous with' to demonst rat e her case, claiming t hat
what woul d be regarded as a ret reat to a more cautious level by relativity theory woul d
be regarded in the reverse light by pre-relativity classical theory. I find t he exampl e
unconvi nci ng on her own terms. But further t han this not e t hat bot h theories agree on
SYSTEMATI C REALI SM 487
t he invariance of simultaneity at the same place among all obs~ver s. It is plausible t o
mai nt ai n t hat this, and not the extended not i on of distant simultaneity, belongs to t he
untheoretical level. That common people may have t hought otherwise is as irrelevant,
in itself, here as for the case of ' sol i d' . )
The empiricists hol d observational reports to be indubitable, I emphat i cal l y do not ,
but hol d t hem open t o theoretically directed criticism. How this can come about is ob-
vious in t he case of those t erms t hat have only pragmatically acquired a reporting rol e
at the behest of theory. Even in t he case of untbeoretical terms, however, we have t o
al l ow t hat t heory is capable of showing these claims t o be false and the corresponding
perceptions illusory (e.g. t he perception of continuity for macr o objects and its semantic
enshrinement in t he concept ' sol i d' ) precisely by explaining how the illusions occur.
For Empiricism t he cat egory of observat i on terms is a single one with no large-scale
epistemic divisions within it. This is equally true of mor e radical Realisms. I distinguish
sharply t he untheoretical level f r om the theory-laden report i ng level within the cat egory
of observat i on terms.
Empiricism holds that all observat i on t erms ar e theory-free and hence t heory neutral
or independent. Radi cal Realisms hol d that all observat i on terms are theory-laden and
hence not t heory neutral. I hol d t hat the terms of the untheoretical level are neutral
vis-a-vis scientific theories, t hough with a theory-like dimension, but that t he terms
t hat have acqui red an observat i onal report i ng rol e ar e t heory laden.
Semantically, observat i on terms ground, or actually exhaust, theoretical terms ac-
cordi ng t o Empiricism. The reverse is the case for terms that have acqui red a report i ng
role in observat i on accordi ng t o my doctrine. Moreover, the degree t o which untheo-
retical terms cont ri but e t o t he semantics of theories is probabl y small, t hough t he
mat t er seems obscure at this t i me (cf. Section 6.6.4.).
For t he foregoing reasons theoretical terms can never genuinely usurp t he rol e of
observat i on terms in Empiricism, for me this is a genuine possibility, t hough it might
requi re considerably different condi t i ons from those now obtaining. In a secondary
sense it has al ready occurred in t he i nt roduct i on of terms with a reporting role.
For according t o my doctrine t he observation reports couched in untheoretical terms
have tittle significance for science, their onl y val ue is that they are needed to supply
part of the instructions for i nt roduci ng theory-laden terms t hat ar e to acqui re a re-
port i ng rol e in observation. This is quite t he opposite of the empiricist rol e for these
terms. For Empi ri ci sm these terms play t he central rol e requi red in a foundat i onal
epistemology, but I espouse a foundationless epistemology.
Because of the foregoing, t he macl o level of common sense observables is the pri mary
level for Empiricism - in most versions of the doctrine it is the only level. In my case it
is the ont ol ogy specified by our best t heory t hat constitutes the pri mary level and the
macr o level is t o be underst ood in terms of it. Nonetheless, unlike many ot her Realisms,
I do not reject the macr o level in tow, for I regard the common sense conceptual frame-
wor k that forms the basis of the untheoretical level as mor e adequat e t han not and
t hat the pereeptual-conceptual discrepancies can be correct ed for t hrough an adequat e
t heory of illusion (perception) and suitable pruni ng of tacit theoretical commi t ment s
(conception).
Finally, I want to add a wor d about the not i on of ostensive definition. (It can, of
course, be called definition in some extended sense of that term.) I have already argued
in a general way for its retention (Section 6.4.1.), now I want t o explicitly retain it in a
limited form. There seems little doubt t hat some concepts ar e particularly closely con-
nected t o specific experiences and derive their linguistic function, at least in significant
488 c. A. HOOKER
part, from this connection. I do not want to say that such concepts have a unique ki nd
of semantical status, much less that their rneanlng is the content of the experience. The
meanings of such concepts are to be spelled out in terms of their total linguistic role
(or whatever), just as for any other term, but in their case the criterion of their appfica-
tion will be fundamentally tied to the (potential) occurrence of the experience and no
explicit definition for them may be possible - only in this sense are these terms se-
mantically segregable. (Notice too that there is nothing to prevent the experience from
being identified in alternative ways.) In this sense I find many terms of the commonsense
vocabulary ostensively definable (and, ff the argument of Section 6.4.1. is to be believed,
necessarily so). These, and my earlier comments on ostensive definitions - see Not e 50 -
undermine Aune' s claim that ff all concepts were to he reducible to ostensively de-
finable ones then one could not avoid behaviourism as a theory of the human per-
sonality, see [8].
7s Despite this general philosophical dispute with Graves, and despite other more tech-
nical criticisms I would make of his account of relativity not pertinent here, I consider
Graves' book specially valuable because it belongs to a rare class of works that attempt
technically competant, yet philosophically sensitive, expositions of a scientific theory.
These days this occurs all too rarely. Yet I believe that in a truly mature and ' alive'
intellectual community all theories would he given such a treatment.
7~ There are ways in which P and Q might be challenged, but they are of no relevance
here. One might e.g. argue that quantum objects are wave packets and i~terfering wave
packets defeat P, but we are discussing compositional relations between levels here -
the table and the corresponding atoms are not interfering with one another. Similarly
we might speculate on 5 or more dimensional manifolds for the world structure, but
here we are talking about ' parallel' 4-dimensional worlds.
so He offers a fourth, namely that it is terribly difficult in fact to relate some theoretical
levels - but this establishes nothing except the youthfulness and probably error of exist-
ing science.
sl The first argument fs that there may be truths at some level that any reasonable re-
ductive mapping would have to transform into falsehoods at other levels. E.g. ' tables
contain no interstices' is a statement generally conceded to he true at the common sense
level but false at the atomic level (cf. pp. 16-20). The second argument is that
. . . the coordination [between levels A and B] is not carried out on a piece-
meal basis, with parts of A identified with the 'corresponding'parts of B in
succession. For an adequate explanation of even a small part of A may in
f act require up to the whole of B; ... (p. 24)
and the third that
... conversely, two phenomena which are quite distinct in ,4 may appeal to
the same parts or aspects of B. (p. 24)
I shall consider these arguments in order.
The first argument is unconvincing on two grounds. First, I do not believe that we
should accept common sense claims so uncritically; a statement which is true for this
level is ' we do not in general see any interstices in tables' and this claim is well ex-
plained by the radiation pattern at the atomic level which indeed is continuous in the
relevant sense. People are constantly claiming too much for their experience, but that
is no reason to burden science with it. Second, even should we find a discrepancy of this
kind we could not so easily avoid, I believe that it is intellectually preferrable to write
it off as mere appearance (i.e. as illusion) than to adopt Graves' alternatives-always
SYSTEMATI C REALI SM 489
assuming t hat an explanation of t he occurrence of the illusion will be fort hcomi ng at
t he mor e fundament al level. Onl y in the severest ~rcumst auces does one counsel radical
courses of act i on such as droppi ng P or Q.
The second argument , and t he third, are also such that i f they were sound they woul d
indeed constitute grounds f or radical change. Unfort unat el y, I can find no explicit
exampl e f or t he second argument in Graves, but this does not prevent a sketch of t he
general er r or on which all such examples are surely based. Consider the phenomenon
of t he r ot at i on of spiral galaxies, and t he specific shape and r ot at i on of our own in
particular. Suppose t hat one want ed an expl anat i on of this phenomenon at the at omi c
level and suppose t hat somet hi ng like Mach' s hypothesis were t rue (i.e. t hat inertia is
a funct i on of all masses i n t he universe). Then t he explanation of this phenomenon
woul d demand reference t o all the entities of t he at omi c level, namel y to all t he atoms.
It is t he plausibility of claims such as these concerning t he expl anat i on of phenomena
t hat lends plausibility t o t he argument. But t he l at t er plausibility is spurious. The ar-
gument is concerned with the ontological distinctness of levels; which entities are in-
vol ved in explanations establishes not hi ng concernin8 ontic distinctness, even where
reductions involving many entities of one level may be needed in t he explanation of
phenomena initially described at t he ot her (simply because many entities at one level
may be involved in t he expl anat i on of phenomena described at t hat very same level
and concerning onl y a subset of these entities)). Our galaxy may still be identical with
a local collection of at oms, t hough all at oms appear in an expl anat i on of its shape and
rot at i on. The appropri at e version of the argument reads: where we have a subset of
t he entities at one level t hat woul d have t o he identified with t he entire ot her level under
any reasonable reduction, then no reductive identification of t he t wo levels is possible.
I woul d accept t he argument but know of no plausible exampl e of t he premise. Only
by sliding between a plausible, but irrelevant, claim and an implausible, t hough rele-
vant, cl ai m does Graves make this argument sound plausible.
The same ' sleight of mi nd' is used t o bolster the third argument, and here Graves
offers us an example. The exampl e is t hat of pp. 21-4, it runs as follows: ' A table corre-
sponds t o a swarm of molecules, a chair also corresponds t o a swarm of molecules,
hence at t he mol ecul ar level tables and chairs are indistinguishable' . The general argu-
ment is: ' Ont ol ogi cal l y distinct things at level A are indistinguishable at level B, so
levels A and B are distinct' . Of course tables and chairs are not merely swarms of mol e-
cules but structured lattices of molecules and the structures may be as readily dis-
tinguished at the molecular, as at t he common sense, level. Indeed, macr o objects ar e
structured, located lattices of molecules, so that even ' i n d i s ~ b l e ' macr o objects
can be readily distinguished at t he mi cr o level. Graves has faked t he argument by a
deliberate omission of specificity in t he premise. Of course, their mol ecul ar structures
do not distinguish tables and chairs qua socially functioning objects, bat then neither
do their macr o physical descriptions which is all that is of Ielevance here (a second
sleight of mi nd in this argument perhaps?). I know of no otheI plausible example for
Gr aves' argument.
Finally, Graves mi ght t ry t o hol d R false, or perhaps meanin$1ess, on the grounds of
incommensurability of theories. (R is an inter-theoretic claim.) I at least woul d reject
this tack, and so, it seems, wotdd Gr a v ~ (of. p. 24).
s2 The idea was to decide upon an opt i mal course of act i on t hat would increase t he
known facts maxi mal l y rapidly and lead by the shortest r out e t o succeeding increasingly
general theories, given t he existing facts and theories. Roughl y speaking this is equi v-
alent t o solving the pr obl em of induction. For i f there is a ' prospect i ve met hodol ogy'
4 9 0 c . A . HOOKE R
(of. my [63]) of discovery which works purely with a dat a-t heory base and churns out ,
by some algorithm, the next best t heory and correlative dat a then it must also thereby
identify the best justified t heory given the new dat a base, and conversely, i f we have a
' ret rospect i ve met hodol ogy' of justification that selects t he best theory, given a data-
t heory base, then it will select the next best t heory in advance given a dat a-t heory base
(or, ff an existing t heory is continually selected, either science has reached its denoue-
ment or met hod becomes blind and one must simply accumulate facts at random).
This narrow construal of met hodol ogy is of course only possible against t he ' logical
machi ne' view of science, for here scientists have been entirely lost sight of. We ignore
the fact that it is a course of act i on for a sci ent i st that is specified, rat her met hodol ogy
is here viewed as provi di ng an al gori t hm for quasi-logical transitions in science.
sa Hanson by at t empt i ng to widen ' t he existing circumstances' t o include a variety of
ot her modes of reasoning t han those admi t t ed by Empiricism. See [44]. Occasionally
Hanson crosses from t he quasi -formal into t he purely psychological, thereby aban-
doni ng his tacit acceptance of the general empiricist account of science. In general
t hough he seems cont ent merely t o enlarge the not i on of formal reasoning wi t h which
Emphi ci sm operated, and the correl at i ve not i on of perception with it.
Feyerabend argues this bot h by at t empt i ng t o reduce t o absurdity the empiricist
met hodol ogy and by argui ng that ' t he existing circumstances' must be enlarged t o in-
clude al l of t he cir~m3stances of an individual' s life.
s4 Taken thus far, there is no necessity t o assert e.g. t hat t heory can onl y be defeated
via confront at i on with anot her t heory (pace Smart ); whether one holds this latter cl ai m
inflexibly or not depends t o a large extent on one' s analysis of t he relations of ' fact ' t o
theory. I f one believed, as Feyerabend seems to, that t he semantics of t he observat i on
language is determined by t he t heory for which it is t he observat i on language t hen the
condi t i on seems at once bot h necessary and impossible t o underst and; for on t he one
hand each theory generates its own real m of ' fact s' on this account and so i f one is t o
know of t he empirical alternatives at all one must first know of the theory, and yet on
the ot her hand since t he ' fact s' of one t heory are semantically i ncommensurat e wi t h
those of t he ot her there seems no way in which to make the direct compari son t hat
empirical j udgment between t hem requires. (Feyerabend' s critics have been qui ck t o
poi nt this out, cf. Achinstein [3], Butts [19], Fi ne [34]; cf. Hesse [55], Hesse [57], Hul l
[76]). Feyerabend' s view, however, was never so consistently, or so simply, drawn. For
one thing he recognizes t he condi t i oned aspect of typical observation reports (this is
his Pragmatic Theory of Observation - cf. [30]); unfortunately he fails to t ake the hint
from Popper' s use of the psychology of perception (as Popper himself fails - of. [101])
and distinguish hard dat a from phenomenal reports. Nonetheless Feyerabend is on
t he right track when he says t hat the level common t o two theories is the level of non-
verbal responses to stimuli (of. the clashes between condi t i oned expectations expressable
at the hard dat a and phenomenal levels of experience), it is j ust t hat wi t hout the
psychological and epistemological apparatus erected earlier one has no way to clearly
understand or developed this intuition.
s5 The changes that have occurred in t he history of science have been quite radical,
even to the ditching of entire ontologies and rafts of ' fact s' = hard data. Only by re-
ducing the theoretical dimension t o near zero can t he empiricist make this history l ook
like a steady accumul at i on of data.
On t he ot her hand, the role of controversy in science is of great i mport ance and it
is implausible to insist that that controversy always amount s to no mor e than propa-
ganda, rat her t han the serious debat e of the comparat i ve merits of different points of
SYSTEMATI C REALI SM 491
view. Fur t her mor e there is considerable evidence for a continuity of progression and
t he preservat i on of fundament al theoretical ideas over the course of science which cannot
be i gnored in any plausible account.
s s Feyerabend at least cannot be said t o be guilty of lack of historical sensitivity.
Indeed, he claims to arri ve at his position t hrough historical analysis. Actually, this is
onl y part l y true, his semantical prescriptions, and t o a large extent his correlative
t heory of observat i on are still formul at ed within the aprioristic mol d.
sv This aspect of t he history of science, of t he relative uni mport ance of mere r aw ob-
servational evidence and the relevant i mport ance of how the t heory dictates what should
be done with it has been nicely stressed by Feyerabend and t oo often ignored by every-
one else.
ss Little enough has been explicitly written recently about systematic metaphysics and
its rel at i on t o science aft er t he fashion of my [72] and [65], t hough I view this as an
urgent necessity. Many of t he earlier writings on this subject, in my view, fall far short
of t he desired perceptiveness and system. Representative among earlier writings are
Burt t [18], Koyr~ [79], [80], Pap [96], Toul mi n and Goodfi el d [132].
se E.g. classical logical structure, and various dynamical structures, f or example t he
canonical dynamical formalism, various group theoretic structures and so on - al t hough
not all of these mat hemat i cal structures are shared by all scientific theories and in some
cases they have been ' general i zed' in a fairly obvious sense. Cf. my [65].
g0 The former usually linked t o the metaphysical-mathematical cont i nui t y level - e.g.
Ei nst ei n' s remarks on t he equivalence of inertia and gravitational mass - and t he l at t er
connected to the el ement ary observational level.
~1 I f a culture presses a particular descriptive scheme and its correl at i ve ont ol ogy over
a l ong peri od of time (e.g. many centuries), constructing within its confines many dif-
ferent successive theories, each one designed t o cope mor e adequately with those areas
of experimental experience where t he former theories r an i nt o difficulty, and i f at t he
end of such a l ong process these difficulties were not clarified but seemed only t o deepen
in their embeddedness in t he fundament al theoretical structure, then it may force t hat
culture t o a re-evaluation of its fundament al descriptive scheme and ont ol ogy and t o
seek t o create alternatives t o it. As a mat t er of fact, I believe t hat we in western so-
ciety are precisely in this position in t he 20th Cent ury where the two great concept ual
schemes ment i oned above have twice met - leading t o t he creation of relativity t heory
and quant um t heory respectively - and the difficulties involved in t he use of those two
schemes, and especially in t he at t empt t o combi ne them, we ar e now discovering, sug-
gests that very fundament al revisions in the basic structure of our description of t he
worl d ar e requi red (cf. [64]). Under these circumstances we may well be forced and
encouraged to l ook around f or alternatives t o these t wo schemes.
I f such be t he case t hen we may wish t o wi t hdraw t he word metaphysics from any
part of t he cont ent of science, withholding t he t erm only f or t he strictly uncriticizable,
but this is really a mat t er of little i mport ance and it is a useful t erm t o draw at t ent i on
t o t he perhaps, but onl y perhaps !, culturally invarlant cont ent of part of our science.
B I B L I OGR AP HY
[1 ] Achinstcin, P., Concepts of Science, Johns Hopki ns Press, Baltimore, 1968.
[2] Achinstein, P., ' On The Meani ng of Scientific Terms' , The Journal of Philosophy
61 (1964), 497-508.
492 c . A. HOO~: eR
[3] Achinstein, P., ' Acute Prolfferitis', in Boston Studies in the Philosophy of Science,
Vol. I I (ed. by R. S. Cohen and M. W. Wartofsky) Humanities Press, New York,
1965.
[4] Achinstein, P. and Barker, S. (eds.), The Legacy of Logical Positivism, Johns
Hopkins Press, Baltimore, 1969.
[5] Agassi, J., ' The Nature of Scientific Problems and Their Roots in Metaphysics',
in The Critical Approach to Science and Philosophy (ed. by M. Buvge), The Free
Press, New York, 1963.
[6] Armstrong, D. M., Perception and the Physical WorM, Routledge & Kegan Paul,
London, 1961.
[7] Aune, B., Knowledge, Mind and Nature, Random House, New York, 1967.
[8] Aune, B,, Rationalism, Empiricism and Pragmatism: An Introduction, Random
House, New York, 1970.
[9] Ayer, A. J., Language, Truth and Logic, Dover, New York, 1946.
[10] Bateson, G., Steps to An Ecology of Mind, Bailanfine, 1972.
[11] Beth, E. W., The Foundations of Mathematics, Harper Torc, hbooks, New York,
1966.
[12] Bohnert, H., ' Communication by Ramsey-Seatence Clause', Philosophy of Sci-
ence 34 (1967), 341-347.
[13] Bohnert, H., ' I n Defense of Ramsey' s Elimination Method' , The Journal of
Philosophy 65 (1968), 275-281.
[14] Braithwaite, R. B., Scientific Explanation, Cambridge Univ. Press, Cambridge,
1959.
[ 15] Bunge, M., Scientific Research, Volumes I & II, Springer-Verlag, New York, 1967.
[16] Bunge, M., ' Theory Meets Experience', publication of the Institut for Veten-
skapsteori, Gt~teborg, August 1971.
[17] Btmge, IV[., The Myth of Simplicity, Engiewood Cliffs, Prentice-Hall New Jersey,
1963.
[18] Burtt, E. A., The Metaphysical Foundations of Modern Physical Science, Rout-
ledge & Kegan Paul, London, 1959.
[19] Butts, R. E., 'Feyerahend and the Pragmatic Theory of Observation', Philosophy
of Science 33 (1966), 383.
[20] Butts, R. E. (ed.), William Whewell's Theory of Scientific Method, Univ. of
Pittsburgh Press, 1968.
[21] Campbell, D. T., 'Methodological Suggestions from a Comparative Psychology
of Knowledge Processes', Inquiry 2 (1959), 152-182.
[22] Campbell, N. R., Foundations of Physics, Dover, New York, 1957.
[23] Carnap, R., TheLogical Structure of the World and Pseudoproblems in Philosophy,
Routledge & Kegan Paul, London, 1967.
[24] Catnap, R., ' The Methodological Character of Theoretical Concepts' in Minne-
sota Studies in the Philosophy of Science, Vol. I (ed. by H. F¢igl and M. Striven),
Univ. of Minnesota Press, Minneapolis, 1956.
[25] Cornman, J. W., ' Craig' s Theorem, Ramsey-scntences and Scientific Instru-
mentalism', Synthese 25 (1972), 82-128.
[26] Duhem, P,, The Aim and Structure of Physical Theory, Atheneum, New York,
1962.
[27] Feyerabend, P. K., 'Problems of Empiricism', in Pittsburgh Studies in the Philos-
ophy of Science, Vol. II (ed. by R. Colodny), Prc~tice-Hall, New Jersey, 1965.
[28] Feyerabend, P. K., 'Problems of Empiricism II' in Pittsburgh Studies in the
SYSTEMATIC REALISM 493
Philosophy of Science, Vol. W (ed. by R. Colodny), Univ. of Pittsburgh Press,
Pittsburgh, 1969.
[29] Feyerabend, E K., ' Das Problem der Existenz Theoretischer Entitaten', in Pro-
bleme der Erkenntnistheorie, Festschrift fiir Viktor Kraft, Vienna 1960.
[30] Feyerabend, P. K., ' An Attempt at a Realistic Interpretation of Experience',
Proceedings of the Aristotelian Society 58 (1958).
[31] Feyerabend, P. K., Knowledge Without Foundation, Oberlin College, mimeo-
graphed, 1961.
[32] Feyerabend, P. K., 'Against Method', in Minnesota Studies in the Philosophy of
Science, Vol. IV (ed. by M. Radner and S. Winokur), Univ. of Minnesota Press,
l~inn~apolis, 1970.
[33] Feyerabend, P. K., ' On the Meaning of Scientific Terms', The Journal of Philo-
sophy 62 {1965), 266.
[34] Fine, A., 'Consistency, Derivability and Scientific Change', The Journal of
Philosophy 62 (1967), 231-240.
[35] Foster, M. H. and Martin, M. L., Probability, Confirmation and Simplicity,
Odyssey Press, New York, 1966.
[36] Fraassen, B. C. van, ' On the Extension of Beth's Semantics of Physical Theories',
Philosophy of Science 37 (1970), 325-339.
[37] Fraassen, B. C. van, ' A Formal Approach to The Philosophy of Science', in
Paradigms and Paradoxes (ed. by R. Colodny), Pittsburgh Studies in the Philo-
sophy of Science, Vol. V, Univ. of Pittsburgh Press, Pittsburgh, 1972.
[38] Frege, G., ' On the Foundations of Geometry' (transl. by M. E. Szabo), Philo-
sophical Review 69 (1960); (first published in 1903).
[39] Giere, R., 'Objective Statistical Methods and Scientific Enquiry', to appear in
the proceedings of The Conference on Foundations of Probability, Statistical
Inference and Statistical Theories of Science, The Univ. of Wcstern Ontario Series
in Philosophy of Science (ed. by W. A. Harper and C. A. Hooker), D. Reidel,
Dordrecht, Holland.
[40] Crraves, J. C., The Conceptual Foundations of Contemporary Relativity Theory,
M.I.T. Press, Cambridge, Mass., 1971.
[41] Crregory, R. L., Eye and Braim" The Psychology of Seeing, McGraw-Hill, World
Univ. Library, New York, 1971.
[42] Cn'finbaum, A., Philosophical Problems of Space and Time, A. A. Knopf, New
York, 1963.
[43] I-~ber, R. N. (¢d.), Information-Processing Approaches to VisualPerception, Holt,
Rinehart and Winston, New York, 1969.
[44] Hanson, N. R.,Patterns of Discovery, Cambridge Univ. Press, Cambridge, Mass.,
1965.
[45] Hanson, N. R., Concept of the Position, Cambridge Univ. Press, Cambridge, 1963.
[46] Harr6, R., The Principles of Scientific Thinking, MacMillan, London, 1970.
[47] Hempel, C., 'Problems and Changes in the Empiricist Criterion of Meaning',
Revue Internationale de Philosophie 11 (1950), 41-63.
[48 ] Hempel, C., 'The Theoretic]art's Dilemma', in Minnesota Studies in the Philosophy
of Science, Vol. II (ed. by H. Feigl and G. Maxwell), Univ. of Minnesota Press,
1958.
[49] Hompel, C., ' On the 'Standard Conception' of Sciontific Theory', in Minnesota
Studies in the Philosophy of Science, Vol. IV. (ed. by IV[. Radner and S.
Winokur), Univ. of Minnesota Press, Minneapolis, 1970.
494 c . A. HOOKER
[50] Hempel, C., Fundamentals of Concept Formation in Empirical Science, The Univ.
of Chicago Press, Chicago, 1952.
[51] Hempel, C., Aspects of Scientific Explanation, The Free Press, New York, 1965.
[52] Henkin, L., Suppes, P., Tarski, A., The Axiomatic Method, North-Holland Publ.
Co., Amsterdam, 1959.
[53] Hesse, M., Models and Analogies in Science, Shoed & Ward, London, 1963.
[54] Hesse, M., Forces and Fields, Littlefield, Adams, New York, 1965.
[55] Hesse, M., ' Fine' s Criteria for Meaning Change', The Journal of PhUosophy 65
(1968), 46-51.
[56] Hesse, M., ' A Self-Correcting Observation Language', in Logic, Methodology
and Philosophy of Science HI (ed. by B. van Rootselaar and J. F. Staal), North-
Holland Publ. Co., Amsterdam, 1968.
[57] Hesse, M., ' Is There an Independent Observation Language?', in The Nature and
Function of Scientific Theories (ed. by R. G. Coloday), Univ. of Pittsburgh
Studies in the Philosophy of Science, Vol. IV, Univ. of Pittsburgh Press, 1970.
[58] Hooker, C. A., ' Secondary Qualities and Systematic Philosophy' , Ph.D. Thesis,
York University, 1970. In preparation as a book.
[59] Hooker, C. A., 'Craigian Tramoriptionism', American Philosophical Quarterly
5 (1968), 1952-163.
[60] Hooker, C. A., ' Five Arguments Against Craigian Transcriptionism', Australa-
sian Journal of Philosophy 46 (1968), 265-276.
[61] Hooker, C. A., ' The Referencial Function of Bound Variables', Mind 80 (1971),
481--498.
[62] Hooker, C. A., 'Critical Notice: Radner, M. and Winokur, S. (eds.), Analyses
of Theories and Methods of Physics and Psychology', Canadian Journal of
Philosophy I (1972), 393--407.
[63] Hooker, C. A., 'Critical Notice: Against Method, P. K. Feyerabend', Canadian
Journal of Philosophy I (1972) 489-509.
[64] Hooker, C. A., ' The Nature of Quantum Mechanical Reality: Einstein Versus
Bohr' , in Paradox and Paradigm, Pittsburgh Studies in the Philosophy of Science
V (ed. by R. G. Colodny), 1972.
[65] Hooker, C. A., 'Metaphysics and ModemPhysics: A Prolegomenon to the Under-
standing of Quantum Theory' , in Contemporary Research in The Foundations
andPhilosophy of Quantum Theory (ed. by C. A. Hooker), Dordrecht, Reidel, 1973.
[66] Hooker, C. A., 'Empiricism, Perceptionand Conceptual Change', Canadian Jour-
nal of Philosophy 3 (1973), 59-75.
[67] Hooker, C. A., ' The Relational Doctrines of Space and Time', The British Jour-
nal of the Philosophy of Science 22 (1971), 97-130.
[68] Hooker, C. A. , 'Defense of a Non-Conventionalist Approach to Classical Me-
chanics', in Boston Studies in the Philosophy of Science XIH, 1974.
[69] Hooker, C. A., ' The Non-Necessity of Qualitative Content', Dialogue 12 (1973).
[70] Hooker, C. A., ' The Philosophy of Paul K. Feyerabend: A Critique', t o be
published.
[71] Hooker, C. A., ' Global Theories', to appear.
[72] Hooker, C. A., ' The Metaphysics of Science: Atoms vs Plena', to appear in
International Logic Review.
[73] Hooker, C. A., ' Has the Scientist any Future in the Brave New World?' , a paper
read at the 500th Anniversary of the Birth of Nicholas Copernicus, Conference,
Ann Arbor, Mich., April 1973. To appear in the proceedings.
SYSTEMATIC REALISM 495
[74] Hooker, C. A., ' Remarks on the Philosophical Uses of the Information-Pro-
cessing Approach to the Brain', not yet published.
[75] Hospers, J., An Introduction to Philosophical Analyses, Englewood Cliffes, Pren-
tice-Hall, New Jersey, 1967.
[76] Hull, R. T., ' Feyerabend' s Attack on Observation Sentences', Synthese 23 (1972),
374-399.
[77] Kohler, I., 'Experiments with Goggles', Scientific American, May, 1962.
[78] K6rner, S., ' On Philosophical Arguments in Physics', in Observation and Inter-
pretation in the Philosophy of Physics (ed. by S. K6rner), Dover, New York,
1962.
[79] Koyr6, A., Metaphysics and Measurement, Cambridge Univ. Press, Cambridge,
Mass., 1908.
[80] Koyr6, A., From Closed Space to Infinite Universe, Johns Hopkins Press, Balti-
more, 1957.
[81 ] Kuhn, T., The Structure of Scientific Revolutions, Univ. of Chicago Press, Chicago,
1966.
[82] Lakatos, I. and Musgrave, A. (eds.), Criticism and the Growth of Knowledge,
Cambridge Univ. Press, Cambridge, 1970.
[83] Leach, J. J. (ed.), Inductive Aspects of Knowledge: Readings in Recent Episte-
mology, Univ. of Western Ontario Series in Philosophy of Science, D. Reidel,
Dordrecht, Holland, to appear.
[84] Lewis, D., ' How to Define Theoretical Terms', The Journal of Philosophy 6 7
(1970), 427--445.
[85] Luckenbach, S., Probabilities, Problems and Paradoxes, Dickenson, California,
1972.
[86] McMullin, E., ' What Do Physical Models Tell Us?' , in Logic, Methodology and
Philosophy of Science HI (ed. by B. van Rootselaar and J. F. Staal), North-
Holland Publ. Co., Amsterdam, 1968.
[87] Marras, A., ' On Sellars' Linguistic Theory of Conceptual Activity', in Canad/an
Journal of Philosophy 2 (1973), 471-484.
[88] Marras, A., 'Conceptual Activity, Rules and Linguistic Actions: A Rejoinder to
Wilfred Sellars', in Canadian Journal of Philosophy 2 (1973), 495-501.
[89] Maxwell, G., ' The Ontological Status of Theoretical Entities', in Minnesota
Studies in the Philosophy of Science, Vol. III (ed. by H. Feigl and G. Maxwell),
Univ. of Minnesota Press, Minneapofis, 1962.
[90] Maxwell, G., 'Structural Realism and the Meaning of Theoretical Terms', in
Minnesota IV, See [32].
[91] Maxwell, G., ' Corroboration Without Demarkation', in The Philosophy of Karl
Popper (ed. by P. S. Schilpp), LaSalle Ill. Open Court, forthcoming.
[92] Nagel, E., The Structure of Science, Routledge & Kegan Paul, London, 1961.
[93] Nagel, T., 'Brain Bisection and the Unity of Consciousness', Synthese 22 (1971),
396-413.
[94] Nash, L. K., The Nature of the Natural Sciences, Little, Brown, Boston, 1963.
[95] Pap, A., An Introduction to the Philosophy of Science, Eyre & Spottiswoode,
London, 1963.
[96] Pap, A., ' Does Science have Metaphysical Presuppositions?', in Readings in the
Philosophy of Science (ed by H. Feigl and M. Brodbeck), Appleton-Century-
Crofts, New York, 1963.
[97] Piaget, J., The Construction of Reality in the Child, Basic Books, New York 1954.
496 c. A. ~OOKER
[98] Piaget, J., Genetic Epistemology (transl. by E. Drakworth), Columbia, NewYork,
1970.
[99] Piaget, J., Biology and Knowledge (transl. by B. Walsh), Univ. of Chicago Press,
Chicago, 1971.
[100] Popper, K. R., The Logic of Scientific Discovery, Hutchinson, London, 1962.
[101] Popper, K. R., 'Philosophy of Science: A Personal Report', in BritishPhiiosophy
in the Mid-Century (ed. by C. A. Mace), Allen and Unwin, London, 1957.
[102] Pribram, K., The Lzmguage of the Brain, Prentice-Hall, Englewood Cliffs, N. J.,
1971.
[103] Putnam, H., ' What Theories are Not' , in Logic, Methodology and Philosophy of
Science (ed. by E. Nagel, P. Suppes, and A. Tarski), Stanford Univ. Press,
Stanford, California, 1962.
[104] Putnam, H., ' How Not to Talk About Meanings' in Boston Studies in the Philo-
sophy of Science, Vol. II (ed. by R. S. Cohen and M. Wartofsky), Humanities
Press, New York, 1965.
[105] Putnam, H., 'The Analytic and the Synthetic', in Minnesota Studies in the Philo-
sophy of Science, Vol. III (ed. by H. Feigl and G. Maxwell), Univ. of Minnesota
Press, Minneapolis, 1962.
[106] Quine, W. V. O., From a LogicalPoint of View, Harper & Row, New York, 1963.
[107] Quine, W. V. O., Word and Object, The M.I.T. Pr~,~s, Cambridge, Mass., 1960.
[108] Quine, W. V. O., Ontological Relativity and Other Essays, Columbia Univ. Press,
New York, 1969.
[109] Ramsey, F. P., The Foundations of Mathematics, Routledge & Kegan Paul,
London, 1965.
[110] Reichenbach, H., The Theory of Relativity and A Priori Knowledge, (transl. by
M. Reichenbach), Univ. of California Press, Los Angeles, 1965.
[111] Rudner, R., ' An Introduction to Simpficity', Philosophy of Science 28 (1961),
109-119.
[112) Russell, B., Human Knowledge: Its Scope and Limits, Simon & Schuster, New
York, 1948.
[113] Russell, B., The Analysis of Matter, Allen & Unwin, London, 1927.
[114] Schaffner, K. F., 'Outlines of a Logic of Comparative Theory Evaluation with
Special Attention to Pre- and Post-Relativistic Electrodynamics', in Historical
and Philosophical Perspectives of Science (ed. by R. H. Stuewer), Minnesota
Studies in the Philosophy of Science, V, Univ. of Minnesota Press, Minneapolis,
1970.
[115] Schick, K., 'Indetermlna¢3 t of Translation', Journal of Philosophy 69 (1972),
818-832.
[116] Sellars, W., Science, Perception and Reality, Routiedge & Kegan Paul, London,
1963.
[117] Sellars, W., 'Scientific Realism or Irenic Instrumentalism', in Boston Studies in
the Philosophy ofScience, Vol. II (ed. by R. S. Cohen and W. Wartofsky), Humani-
ties Press, New York, 1965.
[118] Scllars, W., 'The Identity Approach to the Mind Body Problem', in Boston
Studies in the Philosophy of Science, Vol. II (ed. by B. Cohen and W. Wartofsky),
Humanities Press, New York, 1965.
[119] Sellars, W., 'Reply to Martas', Canadian Journal of Philosophy 2 (1973), 485--494.
[120] Shapere, D., 'Space, Time and Language - An Examination of Some Problems
and Methods of the Philosophy of Science', in Philosophy of Science: The
SYSTEMATIC REALISM 497
Delaware Seminar, Vol. 2, (ed. by B. Baumrin), Interscience, New York, 1963.
[121] Shapere, D., ' Meaning and Scientific Change', in Mind and Cosmos (ed. by R.
G. Colodny), Pittsburgh Studies in the Philosophy of Science, Vol. III, Univ.
of Pittsburgh Press, Pittsburgh, 1966.
[122] Smart, J. J. C., Between Science and Philosophy, Random House, New York,
1968.
[123] Sneed, J. D., The Logical Structure of Mathematical Physics, Humanities Press,
New York, 1971.
[124] Suchting, W., 'Deductive Explanation and Prediction Revisited', Philosophy of
Science 34 (1967), 41-52.
[125] Suppe, F., ' What' s Wrong With the Received View on The Structure of Scientific
Theory?', Philosophy of Science 39 (1972), 1-19.
[126] Suppe, F., 'Theories, Their Formulations, and the Operational Imperative',
Synthese 25 (1972), 129-164.
[127] Suppe, F. (ed.), The Structure of Scientific Theories, Univ. of Illinois Press,
Urbana, Ill., 1973.
[128] Suppe, F., 'Theories and Phenomena', forthcoming in Theory and Decision.
[129] Suppes, P., Studies in the Methodology and Foundations of Science, Humanities
Press, New York, 1969.
[130] Suppes, P., ' What is a Scientific Theory?' , in Philosophy o f Science (ed. by A.
Danto and S. Morgenbesser), World Publishing Co., New York, 1960.
[131] Toulmin, S., Foresight and Understanding, Harper & Row, New York, 1963.
[132] Toulmin, S. and Goodfield, J., The Architecture of Matter, Pelican, London,
1965.
[133] Whewell, W., Novum Organon Renovatum, London, 1858.
[134] Whorf, B. L., Language, Thought, and Reality (ed. by John B. Carroll), MIT
Press, Cambridge, Mass., 1966.
[135] Wilder, H., Toward a Naturalistic Theory of Meaning, Ph. D. Thesis, Univ. of
Western Ontario, March 1973.
[136] Wisdom, J. O., 'Conventionalism, Truth, and Cosmological Furniture' ; a paper
read t o the 1972 Philosophy of Science Association Meetings, Lansing.
[137] Wisdom, J. O., ' Four Contemporary Interpretations of the Nature of Science',
Foundations of Physics I (1971), 269-284.
[138] Wojciechowski, J., ' The Ecology of Knowledge and Major Breakthroughs in
Science'; a paper read t o a ConferenCe celebrating the 500th Anniversary of
Nicholas Copernicus: Science and Society: Past, Present and Future, Ann Arbor,
Michigan, April 1973 - to appear in the proceedings.
[139] Wyburu, C. M., Pickford, R. W. and Hirst, J., The Human Senses and Percep-
tion, Univ. of Toronto, Toronto, 1968.
[140] Ziman, P., Public Knowledge, Cambridge Univ. Press, Cambridge, Mass., 1968.